CN112926952A

CN112926952A - Cloud computing-combined big data office business processing method and big data server

Info

Publication number: CN112926952A
Application number: CN202110354768.9A
Authority: CN
Inventors: 许东俊; 陈宙
Original assignee: Individual
Current assignee: Individual
Priority date: 2021-04-01
Filing date: 2021-04-01
Publication date: 2021-06-08

Abstract

The application discloses a cloud computing-combined big data office business processing method and a big data server, when a real-time office item and a delayed office item are adopted to train a first office item identification strategy, not only the identification result of the first office item identification strategy is considered, but also the identification result of a second office item identification strategy is merged into a real-time office execution progress and a delayed office execution progress, the training process of the first office item identification strategy can be supervised by the identification result of the second office item identification strategy, the first office item identification strategy with higher accuracy and anti-interference performance can be trained, the difference and the relevance among different office items can be fully considered in the office item scheduling process, the rationality of office item scheduling is ensured from the global level, and the normal execution of different office items can be ensured as much as possible, the situation that office matters conflict with each other in the scheduling process is avoided.

Description

Cloud computing-combined big data office business processing method and big data server

Technical Field

The application relates to the technical field of cloud computing and big data office, in particular to a big data office business processing method and a big data server combined with cloud computing.

Background

Cloud computing is a super computing mode based on the internet, thousands of computers and servers are connected into a computer cloud in a remote data center, and users can access the data center through computers, mobile phones and the like and operate according to respective requirements. Cloud computing is a mode of delivery and use of IT infrastructure, and is a way to obtain the required resources in an on-demand, easily scalable manner over a network. Cloud computing rationale: by distributing the computation over a large number of distributed computers, rather than local computers or remote servers, the enterprise data center operates more like the internet.

Generally, cloud computing has the advantages of safe and reliable data, convenience in data sharing, low requirements on clients and the like, and therefore, the cloud computing is gradually involved in various business applications. The combination of cloud computing and big data provides a new processing mode for various business applications, and by taking office business as an example and relying on cloud computing and big data, the current office business is realized on-line and intelligentization, and office cost is effectively saved compared with the traditional office mode.

In practical application, certain office scheduling is generally required for cloud office or remote office to meet different office business requirements, but the related office scheduling technology still has certain defects.

Disclosure of Invention

In order to solve the technical problems in the related art, the application provides a big data office business processing method and a big data server combined with cloud computing.

The application provides a big data office business processing method combined with cloud computing, which comprises the following steps:

acquiring a first real-time office item, a real-time office item corresponding to the first real-time office item, a first state office item and a second state office item, wherein the first state office item is obtained by performing office item identification on the first real-time office item based on a first office item identification strategy, and the second state office item is obtained by performing office item identification on the first real-time office item based on a second office item identification strategy;

determining a first real-time office execution progress based on the real-time office matters corresponding to the first real-time office matters, the first state office matters and the second state office matters;

acquiring a delayed office item, a third state office item and a fourth state office item, wherein the third state office item is obtained by performing office item identification on the delayed office item based on the first office item identification strategy, and the fourth state office item is obtained by performing office item identification on the delayed office item based on the second office item identification strategy;

determining a first delayed office execution progress based on the third status office transaction and the fourth status office transaction;

and training the first office item identification strategy based on the first real-time office execution progress and the first delayed office execution progress, and performing office item scheduling processing based on the trained office item identification strategy.

The application also provides a big data server, which comprises a processor and a memory; the processor is connected with the memory in communication, and the processor is used for reading the computer program from the memory and executing the computer program to realize the method.

The cloud-computing-combined big data office business processing method and the big data server provided by the embodiment of the application, when the real-time office matters and the delayed office matters are adopted to train the first office matter identification strategy, not only the identification result of the first office matter identification strategy is considered, but also the identification result of the second office matter identification strategy is integrated into the real-time office execution progress and the delayed office execution progress, so that the first office matter identification strategy learns the identification result of the second office matter identification strategy, therefore, the training process of the first office matter identification strategy can be supervised by the identification result of the second office matter identification strategy, the first office matter identification strategy with higher accuracy and anti-interference performance can be trained, the difference and relevance among different office matters can be fully considered in the office matter scheduling process, therefore, the rationality of office item scheduling is ensured from the global level, the normal execution of different office items is ensured as far as possible, and the condition that the office items conflict with each other in the scheduling process is avoided.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and together with the description, serve to explain the principles of the application.

Fig. 1 is a schematic hardware structure diagram of a big data server according to an embodiment of the present disclosure.

Fig. 2 is a flowchart of a big data office business processing method combining cloud computing according to an embodiment of the present application.

Fig. 3 is a block diagram of a big data office business processing device incorporating cloud computing according to an embodiment of the present application.

Fig. 4 is a schematic communication architecture diagram of a big data office business processing system combined with cloud computing according to an embodiment of the present application.

Detailed Description

Reference will now be made in detail to the exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, like numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present application, as detailed in the appended claims.

It should be noted that the terms "first," "second," and the like in the description and claims of this application and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order.

The method provided by the embodiment of the application can be executed in a big data server, a computer device or a similar operation device. Taking the operation on a big data server as an example, fig. 1 is a hardware structure block diagram of the big data server implementing the big data office business processing method combined with cloud computing according to the embodiment of the present application. As shown in fig. 1, the big data server 10 may include one or more processors 102 (only one is shown in fig. 1) (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA) and a memory 104 for storing data, and optionally, the big data server 10 may further include a transmission device 106 for communication function. It will be understood by those skilled in the art that the structure shown in fig. 1 is merely illustrative and is not intended to limit the structure of the big data server 10. For example, big data server 10 may also include more or fewer components than shown in FIG. 1, or have a different configuration than shown in FIG. 1.

The memory 104 may be used to store computer programs, for example, software programs and modules of application software, such as a computer program corresponding to the big data office business processing method combined with cloud computing in the embodiment of the present application, and the processor 102 executes various functional applications and data processing by running the computer programs stored in the memory 104, that is, implements the method described above. The memory 104 may include high speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include memory located remotely from the processor 102, which may be connected to the big data server 10 over a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.

The transmission device 106 is used for receiving or transmitting data via a network. The above-described specific examples of the network may include a wireless network provided by a communication provider of the big data server 10. In one example, the transmission device 106 includes a Network adapter (NIC), which can be connected to other Network devices through a base station so as to communicate with the internet. In one example, the transmission device 106 may be a Radio Frequency (RF) module, which is used for communicating with the internet in a wireless manner.

It will be understood that the terms "first," "second," and the like as used herein may be used herein to describe various concepts, which are not limited by these terms unless otherwise specified. These terms are only used to distinguish one concept from another. For example, a first real-time office event may be referred to as a second real-time office event, and similarly, the second real-time office event may be referred to as the first real-time office event, without departing from the scope of the present application.

For example, the at least one real-time office event may be any integer number of real-time office events greater than or equal to one, such as one real-time office event, two real-time office events, three real-time office events, and the like. The plurality means two or more, and for example, the plurality of real-time office events may be an integer number of real-time office events of two or more, such as two real-time office events and three real-time office events. Each refers to each of the at least one, for example, each of the plurality of real-time office events refers to each of the plurality of real-time office events, and if the plurality of real-time office events is three real-time office events, each of the plurality of real-time office events refers to each of the three real-time office events.

First, a brief description will be given of the related art adopted in the present application.

(1) Artificial Intelligence (AI) is a theory, method, technique and application system that uses a digital computer or a machine controlled by a digital computer to simulate, extend and expand human Intelligence, perceive the environment, acquire knowledge and use the knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technique of computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can react in a manner similar to human intelligence. Artificial intelligence is the research of the design principle and the realization method of various intelligent machines, so that the machines have the functions of perception, reasoning and decision making.

(2) The artificial intelligence technology is a comprehensive subject and relates to the field of extensive technology, namely the technology of a hardware level and the technology of a software level. The artificial intelligence infrastructure generally includes technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technologies, operation/interaction systems, mechatronics, and the like. Artificial intelligence software techniques include natural language processing techniques and machine learning.

(3) Machine Learning (ML) is a multi-domain cross discipline, and relates to a plurality of disciplines such as probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory and the like. The special research on how a computer simulates or realizes the learning behavior of human beings so as to acquire new knowledge or skills and reorganize the existing knowledge structure to continuously improve the performance of the computer. Machine learning is the core of artificial intelligence, is the fundamental approach for computers to have intelligence, and is applied to all fields of artificial intelligence. The machine learning and the deep learning comprise technologies such as artificial neural network, belief network, reinforcement learning, transfer learning, inductive learning, teaching learning and the like.

(4) The cloud computing technology is a hosting technology for unifying series resources such as hardware, software, networks and the like in a wide area network or a local area network to realize the computation, storage, processing and sharing of data. The cloud technology is a general term of a network technology, an information technology, an integration technology, a management platform technology, an application technology and the like based on cloud computing application, can form a resource pool, is used as required, and is flexible and convenient. Background services of the technical network system require a large amount of computing and storage resources, such as video websites, picture websites and more portal websites, and the cloud computing technology becomes an important support. With the high development and application of the internet industry, each article may have its own identification mark and needs to be transmitted to a background system for logic processing, data of different levels are processed separately, and various industrial data need strong system background support and can only be realized by a cloud computing technology. As a basic capability provider of cloud computing, a cloud computing resource pool, referred to as an IaaS (Infrastructure as a Service) platform for short, is established, and multiple types of virtual resources are deployed in the resource pool and are used by external clients selectively.

The big data office business processing method combining cloud computing provided by the embodiment of the application will be described below based on an artificial intelligence technology and a cloud technology.

The embodiment of the application provides a big data office business processing method combined with cloud computing, and an execution main body is a big data server.

In one possible implementation, the big data server is a smart device, and the smart device may be a plurality of types of smart devices, such as, but not limited to, a smart phone, a tablet computer, a notebook computer, or a desktop computer.

In another possible implementation manner, the big data server is a server, where the server is an independent physical server, or a server cluster or a distributed system formed by a plurality of physical servers, or a cloud server that provides basic cloud computing services such as a cloud service, a cloud database, cloud computing, a cloud function, cloud storage, a network service, cloud communication, a middleware service, a domain name service, a security service, a CDN (content delivery network), and a big data and artificial intelligence platform, and the application is not limited herein.

On the basis, fig. 2 is a flowchart of a big data office business processing method combining cloud computing according to an embodiment of the present application. The execution subject of the embodiment of the present application is a big data server, and referring to fig. 2, the method may include the following:

step 101, acquiring a first real-time office event, a real-time office event corresponding to the first real-time office event, a first-state office event and a second-state office event.

In the embodiment of the application, the office item identification refers to identifying an office interaction node in an office item, marking whether the office interaction node belongs to a target identification node, and obtaining a state office item corresponding to the office item, so as to identify content information of the target office item in the office item. The real-time office event has a corresponding real-time office event, and in some optional embodiments, the real-time office event is obtained by marking the real-time office event by the user. The delayed office event lacks a corresponding delayed status office event. Because the delayed office event lacks a corresponding delayed office event, a policy (artificial intelligence model) capable of identifying the delayed office event cannot be trained based on the delayed office event.

The real-time office item and the delayed office item have difference information, and different standards can be adopted to identify the real-time office item and the delayed office item according to requirements in the embodiment of the application. For example, the collection mode of the real-time office event is different from that of the delayed office event, the real-time office event is the office event corresponding to the current office node collected by the first office event collection device, and the delayed office event is the office event corresponding to the current office node collected by the second office event collection device. For another example, the real-time office event is different from the event completion condition corresponding to the delayed office event, the real-time office event includes completed office events, and the delayed office event includes incomplete office events. For another example, the feature labels of the real-time office event and the delayed office event are different, the real-time office event is a global event set, and the delayed office event is a local event set.

Therefore, in the embodiment of the present application, a model training method based on an office state adaptive mechanism is provided for distribution difference information between real-time office matters and delayed office matters, and an office matter identification strategy (office matter identification model) for office matter identification is trained without supervision in a manner that two models learn and supervise each other.

The first office affair recognition strategy and the second office affair recognition strategy are two different models of countertraining, and the first office affair recognition strategy and the second office affair recognition strategy are used for recognizing and processing office affairs to obtain corresponding state office affairs. In some optional embodiments, the model network layer distribution of the first office item identification policy and the second office item identification policy is the same, or the model network layer distribution of the first office item identification policy and the second office item identification policy is different, which is not limited in this application embodiment.

The first office affair identification strategy and the second office affair identification strategy are models for identifying delayed office affairs, and in the process of training the first office affair identification strategy, the big data server acquires the first real-time office affairs, the real-time office affairs corresponding to the first real-time office affairs, the first state office affairs and the second state office affairs. The real-time office event corresponding to the first real-time office event is a global state office event corresponding to the first real-time office event, the first state office event is obtained by performing office event identification on the first real-time office event based on a first office event identification strategy in the current training turn, the second state office event is obtained by performing office event identification on the first real-time office event based on a second office event identification strategy in the previous training turn, and the second state office event is also called a potential state office event corresponding to the first real-time office event.

Step 102, determining a first real-time office execution progress based on the real-time office affairs, the first state office affairs and the second state office affairs corresponding to the first real-time office affairs.

After the big data server determines the real-time office item, the first state office item and the second state office item, a first real-time office execution progress of the first state office item is determined based on the difference information between the real-time office item and the first state office item and the difference information between the second state office item and the first state office item.

And 103, acquiring delayed office events, third-state office events and fourth-state office events.

In the process of training the first office event identification strategy, the big data server also acquires a delayed office event, a third state office event and a fourth state office event. The delayed office items are different from the items corresponding to the real-time office items in the completion condition, the third-state office items are obtained by performing office item identification on the delayed office items based on the first office item identification strategy in the current training turn, the fourth-state office items are obtained by performing office item identification on the delayed office items based on the second office item identification strategy in the previous training turn, and the fourth-state office items are also called potential-state office items corresponding to the delayed office items.

And 104, determining a first delayed office execution progress based on the third state office item and the fourth state office item.

After the big data server determines the third status office event and the fourth status office event, the first delayed office execution progress of the third status office event is determined based on the difference information between the third status office event and the fourth status office event.

And 105, training a first office item identification strategy based on the first real-time office execution progress and the first delayed office execution progress.

After the big data server determines the first real-time office execution progress and the first delayed office execution progress, based on the first real-time office execution progress and the first delayed office execution progress, model parameters of the first office item identification strategy are adjusted, so that disturbance variable quantities of the first real-time office execution progress and the first delayed office execution progress are less and less until the first real-time office execution progress and the first delayed office execution progress tend to be stable, and the trained first office item identification strategy is obtained.

And step 106, performing office item scheduling processing based on the trained office item identification strategy.

The office item identification strategy obtained through training has high accuracy and anti-interference performance, the big data server can perform office item scheduling processing on any office item based on the office item identification strategy, and differences and relevance among different office items can be fully considered in the office item scheduling process, so that the rationality of office item scheduling is ensured on the global level, normal execution of different office items is ensured as far as possible, and the situation that office items conflict with each other in the scheduling process is avoided.

The method provided by the embodiment of the application, when the real-time office matters and the delayed office matters are adopted to train the first office matter identification strategy, not only the identification result of the first office matter identification strategy is considered, but also the identification result of the second office matter identification strategy is merged into the real-time office execution progress and the delayed office execution progress, so that the first office matter identification strategy learns the identification result of the second office matter identification strategy, the training process of the first office matter identification strategy can be supervised by the identification result of the second office matter identification strategy, the first office matter identification strategy with higher accuracy and anti-interference performance can be trained, the difference and the relevance among different office matters can be fully considered in the office matter scheduling process, and the rationality of office matter scheduling can be ensured in the global aspect, the normal execution of different office matters is ensured as far as possible, and the condition that the office matters conflict with each other in the scheduling process is avoided.

Based on the above, the following shows a big data office business processing method combining cloud computing provided by the embodiment of the present application, where an execution subject of the embodiment of the present application is a big data server, and the method may include the following steps:

step 201, the big data server obtains a first real-time office event, a real-time office event corresponding to the first real-time office event, a first state office event and a second state office event.

In the process of training the first office event identification strategy, the big data server acquires the first real-time office event, the real-time office event corresponding to the first real-time office event, the first state office event and the second state office event. The real-time office event corresponding to the first real-time office event is a global state office event corresponding to the first real-time office event, the first state office event is obtained by identifying the office event of the first real-time office event based on a first office event identification strategy in the current network training round, the first state office event can be called a predicted state office event corresponding to the first real-time office event, the second state office event is obtained by identifying the office event of the first real-time office event based on a second office event identification strategy in the last training round, and the second state office event is also called a potential state office event corresponding to the first real-time office event.

In some alternative embodiments, the first real-time office event is any type of office event. For example, if the first real-time office event is an office event corresponding to the current office node, the office event corresponding to the current office node is subjected to office event identification processing, and the processed office content or unprocessed office content is identified from the office event corresponding to the current office node. In some optional embodiments, the first real-time office event and the real-time status office event corresponding to the first real-time office event are obtained by the big data server from a local database, or are obtained by the big data server through downloading from a network, which is not limited in this application.

In the embodiment of the application, the first office affair recognition strategy and the second office affair recognition strategy are synchronously trained in a countervailing training mode. The model network layer distribution of the first office affair identification strategy and the second office affair identification strategy is the same, or the model network layer distribution of the first office affair identification strategy and the second office affair identification strategy is different. Model network layer distribution and model parameters of the first office item identification strategy and the second office item identification strategy are set and adjusted according to requirements.

In one possible implementation, the model network layer distributions and model parameters of the first office item identification strategy and the second office item identification strategy are different, and the two strategies can be trained and learned from different angles due to the informativeness of the difference between the model network layer distributions and the model parameters of the two strategies, so that different identification strategy methods are generated, namely, different learning capabilities are provided, and therefore, synchronous supervision among the models can be promoted.

In some alternative embodiments, the first office transaction identification policy is an office transaction identification policy with a feedback-type network as the primary framework). In some alternative embodiments, the second office event identification strategy is based on a lightweight network, thereby reducing the parametric size and operational load of the model.

In step 202, the big data server determines a first real-time office execution progress based on the real-time office event, the first status office event and the second status office event when the first real-time office event is determined to be an abnormal real-time office event based on the second office event identification policy.

In the last training turn, the big data server identifies the office items of the first real-time office item based on the second office item identification strategy to obtain the status office item, determines the real-time office execution progress based on the global real-time office item corresponding to the first real-time office item, the currently obtained status office item and the potential status office item corresponding to the first real-time office item during the training of the second office item identification strategy, and determines the first real-time office item as the abnormal real-time office item based on the real-time office execution progress. The potential state office affairs corresponding to the first real-time office affairs during the training of the second office affairs identification strategy refer to state office affairs obtained by performing office affairs identification on the first real-time office affairs in the previous training turn by the first office affairs identification strategy. The process of determining the abnormal real-time office event based on the real-time office execution progress is detailed in the following step 215, and will not be described here.

The big data server determines a first real-time office execution progress based on the real-time status office event, the first status office event, and the second status office event, in a case where the first real-time office event is determined to be an abnormal real-time office event.

In one possible implementation manner, the big data server determines a third real-time office execution progress based on the real-time office matters and the first state office matters, determines a fourth real-time office execution progress based on the second state office matters and the first state office matters, and determines the first real-time office execution progress based on the third real-time office execution progress and the fourth real-time office execution progress.

The big data server determines a third real-time office execution progress based on the difference information between the real-time office item and the first state office item, and determines a fourth real-time office execution progress based on the difference information between the second state office item and the first state office item. In some optional embodiments, the big data server performs fusion based on a time-sequence layer on the third real-time office execution progress and the fourth real-time office execution progress to obtain the first real-time office execution progress.

In another possible implementation manner, if the first status office transaction includes a first status corresponding to each office interaction node in the first real-time office transaction, determining a third real-time office execution schedule includes: and the big data server determines node state weight corresponding to each office interaction node based on the real-time state office items, and determines a third real-time office execution progress based on the first state and the node state weight corresponding to each office interaction node. In some optional embodiments, the node state weights corresponding to the plurality of office interaction nodes are represented in a node state weight characteristic diagram manner.

In some optional embodiments, determining the node state weight comprises: the big data server determines the minimum progress completion difference between each office interactive node and the progress state value of the real-time state office item, determines the maximum value of the determined minimum progress completion differences as a target value, and determines the node state weight corresponding to each office interactive node respectively based on the target value and the minimum progress completion difference corresponding to each office interactive node.

The maximum value of the minimum difference values of the progress completion is determined as the target value by the big data server, and the node state weight corresponding to each office interaction node is determined based on the difference information between the minimum difference values of the progress completion and the target value.

In some optional embodiments, the process of determining the execution progress of the third real-time office includes: the big data server determines a first office state comparison result based on a first state and a node state weight corresponding to each office interactive node, determines a third global real-time execution progress corresponding to any office interactive node based on the first office state comparison result, the first state and the node state weight corresponding to any office interactive node, and determines a third real-time office execution progress based on the determined multiple third global real-time execution progresses.

Wherein the first office state comparison result represents difference information between the first state office event and the real-time state office event. After the big data server determines the first office state comparison result, a third global real-time execution progress corresponding to any office interactive node can be determined based on the first office state comparison result, the first state corresponding to any office interactive node and the node state weight, so that third global real-time execution progresses corresponding to a plurality of office interactive nodes are determined, fusion based on a time sequence layer is carried out on the determined third global real-time execution progresses, and the third real-time office execution progress is obtained.

In another possible implementation, the big data server trains the first office event recognition strategy in a multiple network parameter optimization manner, since as the number of network parameter optimizations increases, the first office transaction identification strategy is more and more stable, so the big data server adjusts the weight updating frequency corresponding to the node state weight of the third real-time office execution progress and the fourth real-time office execution progress according to the network parameter optimization times, the weight updating frequency corresponding to the node state weight of the fourth real-time office execution progress is increased along with the increase of the network parameter optimization times, thereby avoiding the training process from being affected by erroneous potential status office events generated by unstable office event recognition strategies, meanwhile, effective information in abnormal state office matters is utilized, and stability of office matter identification strategies is enhanced. Determining the first real-time office execution progress based on the third real-time office execution progress and the fourth real-time office execution progress includes: the big data server obtains the network parameter optimization times corresponding to the current network training, responds to the fact that the network parameter optimization times are not smaller than the first set times and not larger than the second set times, and determines the first real-time office execution progress based on the third real-time office execution progress, the fourth real-time office execution progress, the network parameter optimization times, the first set times and the second set times. Or, in response to the network parameter optimization times being greater than the second set times, determining the first real-time office execution progress based on the third real-time office execution progress and the fourth real-time office execution progress.

In some optional embodiments, in response to the number of times of network parameter optimization being less than the first set number of times, the big data server directly determines the third real-time office execution progress as the first real-time office execution progress, without considering the fourth real-time office execution progress.

Step 203, the big data server obtains the second real-time office item, the real-time office item corresponding to the second real-time office item, and the fifth real-time office item.

In the process of training the first office event identification strategy, the big data server also acquires the second real-time office event, the real-time office event corresponding to the second real-time office event and the fifth state office event. The second real-time office item is an office item used for training the second office item identification strategy in the previous training round, the real-time state office item corresponding to the second real-time office item is a global state office item corresponding to the second real-time office item, and the fifth state office item is obtained by performing office item identification on the second real-time office item based on the first office item identification strategy in the current training round.

In the embodiment of the present application, the first real-time office event is an office event of the real-time office event determined to be abnormal based on the second office event identification policy, and the second real-time office event is an office event of the real-time office event determined to be normal based on the second office event identification policy.

In step 204, the big data server determines a second real-time office execution progress based on the real-time office item and the fifth real-time office item corresponding to the second real-time office item when the second real-time office item is determined to be the normal real-time office item based on the second office item identification policy.

Similarly to the process of determining the first real-time office event as the abnormal real-time office event in step 202, in step 204, the big data server determines a second real-time office execution schedule based on the real-time office event and the fifth real-time office event corresponding to the second real-time office event, when the second real-time office event is determined as the normal real-time office event.

In one possible implementation, the real-time status office event includes a real-time status corresponding to each office interaction node in the second real-time office event, and the fifth status office event includes a fifth status corresponding to each office interaction node in the second real-time office event. Determining the second real-time office execution progress comprises: determining a third office state comparison result based on the real-time state and the fifth state corresponding to each office interaction node; determining a second global real-time execution progress corresponding to any office interaction node based on the third office state comparison result, the real-time state corresponding to any office interaction node and the fifth state; determining a second real-time office execution schedule based on the determined plurality of second global real-time execution schedules.

Wherein the third office state comparison result represents difference information between the real-time office event and the fifth office event. After the big data server determines a third office state comparison result, a second global real-time execution progress corresponding to any office interactive node can be determined based on the third office state comparison result and the real-time state and the fifth state corresponding to the office interactive node, so that second global real-time execution progresses corresponding to a plurality of office interactive nodes are determined, fusion based on a time sequence layer is carried out on the determined second global real-time execution progresses, and the second real-time office execution progress is obtained.

In one possible implementation manner, before determining the second real-time office execution progress, the big data server first obtains the sixth status office item, determines a similarity between the fifth status office item and the sixth status office item, and determines the second real-time office execution progress based on the real-time status office item and the fifth status office item corresponding to the second real-time office item in response to the similarity not being less than a preset similarity threshold. The sixth status office event is obtained by performing office event recognition on the second real-time office event based on the second office event recognition strategy.

Since the second real-time office event is a normal real-time office event determined based on the second office event identification policy, in order to further determine whether the second real-time office event is an abnormal real-time office event or a normal real-time office event, the big data server determines a similarity between an identification result of the first office event identification policy for the second real-time office event and an identification result of the second office event identification policy for the second real-time office event, that is, determines a similarity between the fifth status office event and the sixth status office event.

If the similarity is not less than the preset similarity threshold, it is indicated that the second real-time office item is indeed a normal real-time office item, and a second real-time office execution progress corresponding to the second real-time office item is determined by adopting a processing mode of the normal real-time office item. Therefore, in response to the similarity being not smaller than the preset similarity threshold, the big data server determines a second real-time office execution progress based on the real-time office item and the fifth real-time office item corresponding to the second real-time office item.

And if the similarity is smaller than a preset similarity threshold, re-determining the second real-time office item as an abnormal real-time office item, and determining a second real-time office execution progress corresponding to the second real-time office item by adopting a processing mode of the abnormal real-time office item. Therefore, in response to the similarity being smaller than the preset similarity threshold, the big data server determines a second real-time office execution progress based on the real-time office item, the fifth state office item and the sixth state office item corresponding to the second real-time office item. In the case that the similarity is smaller than the preset similarity threshold, the manner of determining the second real-time office execution progress by the big data server is the same as the manner of determining the first real-time office execution progress in step 202, and details are not repeated here.

In some optional embodiments, the big data server determines a dynamic office execution schedule between the fifth status office event and the sixth status office event, and determines the second real-time office execution schedule based on the real-time status office event, the fifth status office event and the sixth status office event corresponding to the second real-time office event in response to the dynamic office execution schedule not being less than the preset execution schedule because the greater the dynamic office execution schedule between the fifth status office event and the sixth status office event, the smaller the similarity between the fifth status office event and the sixth status office event. Or the big data server responds that the dynamic office execution progress is smaller than the preset execution progress, and the second real-time office execution progress is determined based on the real-time office item and the fifth state office item corresponding to the second real-time office item.

In the embodiment of the present application, when the real-time office event is determined to be a normal real-time office event, the method in steps 203 to 204 is adopted to determine the real-time office execution progress corresponding to the normal real-time office event based on the global status office event corresponding to the normal real-time office event. In the case that the real-time office event is determined to be an abnormal real-time office event, the method in step 201 and 202 is adopted to determine the real-time office execution progress corresponding to the abnormal real-time office event based on the global status office event and the potential status office event corresponding to the abnormal real-time office event. When the real-time office event is determined to be an abnormal real-time office event, the global state office event corresponding to the abnormal real-time office event is not accurate enough, so that the potential state office event obtained by introducing the second office event identification strategy to identify the office event of the abnormal real-time office event is supervised in the training process of the first office event identification strategy through the identification result of the second office event identification strategy, errors caused by the abnormal real-time office event in the training process can be reduced, and the anti-interference performance of the office event identification strategy on abnormal data is improved.

Step 205, the big data server obtains the delayed office event, the third status office event and the fourth status office event.

In the process of training the first office event identification strategy, the big data server also acquires a delayed office event, a third state office event and a fourth state office event. The third-state office event is obtained by performing office event identification on the delayed office event based on the first office event identification strategy in the current training round, the fourth-state office event is obtained by performing office event identification on the delayed office event based on the second office event identification strategy in the previous training round, and the fourth-state office event is also called a potential-state office event corresponding to the delayed office event.

The real-time office item and the delayed office item can be identified by adopting different standards according to requirements. For example, a real-time office event is one that includes completed office events, and a delayed office event is one that includes incomplete office events. Alternatively, the real-time office event is an office event corresponding to the current office node collected by the office event collecting device d1, and the delayed office event is an office event corresponding to the current office node collected by the office event collecting device d 2.

In step 206, the big data server determines a first delayed office execution progress based on the third status office item and the fourth status office item.

After the big data server acquires the third state office item and the fourth state office item, the first delayed office execution progress is determined based on the difference information between the third state office item and the fourth state office item.

In one possible implementation manner, the third status office event includes a third status corresponding to each office interaction node in the delayed office event, and the fourth status office event includes a fourth status corresponding to each office interaction node in the delayed office event, and then determining the first delayed office execution progress includes: determining a second office state comparison result based on the third state and the fourth state corresponding to each office interaction node, and determining a local delay execution progress corresponding to any office interaction node based on the second office state comparison result, the third state and the fourth state corresponding to any office interaction node; and determining a first delayed office execution progress based on the determined plurality of local delayed execution progresses.

Wherein the second office state comparison result represents difference information between the third state office event and the fourth state office event. After the big data server determines the second office state comparison result, the local delay execution progress corresponding to any office interaction node can be determined based on the second office state comparison result and the third state and the fourth state corresponding to the office interaction node, so that the local delay execution progress corresponding to the office interaction nodes is determined, the determined local delay execution progress is fused based on a time sequence layer, and the first delay office execution progress is obtained.

And step 207, the big data server trains a first office item identification strategy based on the first real-time office execution progress, the second real-time office execution progress and the first delayed office execution progress.

The big data server adjusts parameters of the first office item identification strategy based on the first real-time office execution progress, the second real-time office execution progress and the first delayed office execution progress, so that disturbance variable quantities of the first real-time office execution progress, the second real-time office execution progress and the first delayed office execution progress are less and less, until the first real-time office execution progress, the second real-time office execution progress and the first delayed office execution progress tend to be stable, and the trained first office item identification strategy is obtained.

In one possible implementation manner, the big data server determines and identifies the office execution progress based on the first real-time office execution progress, the second real-time office execution progress and the first delayed office execution progress, and trains the first office item identification strategy based on the identified office execution progress so that the identified office execution progress tends to be stable.

In the embodiment of the present application, the process of training the first office event recognition policy will be described by taking only the case where the first real-time office event is determined as an abnormal real-time office event and the case where the second real-time office event is determined as a normal real-time office event as an example. In another embodiment, the big data server may not perform steps 203-204 described above, and in step 202, determine a first real-time office execution schedule directly based on the real-time office event, the first status office event, and the second status office event, and then train the first office event identification policy based on the first real-time office execution schedule and the first delayed office execution schedule.

It should be noted that, in the above steps 201 to 207, only one first real-time office event, one second real-time office event, and one delayed office event are processed to obtain a first real-time office execution schedule, a second real-time office execution schedule, and a first delayed office execution schedule. In another embodiment, the big data server processes the first real-time office events, the second real-time office events and the delayed office events, determines the first real-time office execution schedules, the second real-time office execution schedules and the first delayed office execution schedules by the method provided in the above steps 201 to 207, and trains the first office event recognition strategy based on the first real-time office execution schedules, the second real-time office execution schedules and the first delayed office execution schedules.

And step 208, the big data server performs office item detection processing on the first state office item and the third state office item respectively based on the office item detection model to obtain a real-time detection result and a delay detection result.

The big data server performs office item identification processing on the first real-time office item and the delayed office item based on the first office item identification strategy to obtain a first state office item and a third state office item, performs office item detection processing on the first state office item based on the office item detection model to obtain a real-time detection result, and performs office item detection processing on the third state office item based on the office item detection model to obtain a delayed detection result.

The office affair detection model is used for detecting whether the input state office affair belongs to the state office affair of the real-time office affair or the state office affair of the delayed office affair, and can integrate the state office affair of the real-time office affair and the delayed office affair into the counterwork learning process. When the office item detection model performs office item detection processing on the third state office item corresponding to the delayed office item and the obtained delay detection result indicates that the third state office item is the state office item of the delayed office item, it indicates that the office item detection model is accurate in determination, the correlation between the state office item of the delayed office item output by the first office item identification strategy and the state office item of the real-time office item in the feature space is low, and the first office item identification strategy cannot bypass the office item detection model, and the conversion from real-time to delay is not completed, so that the accuracy of the first office item identification strategy in performing office item identification on the delayed office item is still low. When the office item detection model performs office item detection processing on the third state office item corresponding to the delayed office item and the obtained delay detection result indicates that the third state office item is the state office item of the real-time office item, it indicates that the office item detection model is in error, that is, the correlation between the state office item of the delayed office item output by the first office item identification policy and the state office item of the real-time office item is higher in the feature space, and the first office item identification policy completes the conversion from real-time to delay, so that the accuracy of office item identification on the delayed office item by the first office item identification policy is higher and the anti-interference capability is strong.

And step 209, the big data server determines an office execution progress detection result based on the real-time detection result, the delay detection result and the third state office item.

In one possible implementation manner, the third status office event includes a third status corresponding to a plurality of office interaction nodes in the delayed office event, and determining the office execution progress detection result includes: the big data server determines a matter skipping result corresponding to the office matters in the third state based on the third state corresponding to the office interactive nodes; and determining an office execution progress detection result based on the item jump result, the real-time detection result and the delay detection result.

In some optional embodiments, the item skip result corresponding to the third-state office item is used to identify a degree of dynamic change of the third-state office item. And introducing an item heat change value into each office interaction node of the delayed office item, and then determining an office execution progress detection result based on the item heat change value and a detection result output by the office item detection model, so that the heat node state weight of the office interaction node with dynamic change can be increased, and the heat node state weight of the deterministic office interaction node can be reduced. Under the drive of the item heat degree change value, the method is beneficial to enabling the office item detection model to learn how to locate the content with the discrimination.

Step 210, the big data server trains a first office affair recognition strategy and an office affair detection model based on the office execution progress detection result.

After the big data server determines an office execution progress detection result, the network parameter of the first office item identification strategy and the network parameter of the office item detection model are adjusted based on the office execution progress detection result, so that the state office items obtained by identifying the office items of the real-time office items and the delayed office items by the first office item identification strategy are more and more similar in content description, and the office item detection model cannot more and more distinguish whether the state office items are the state office items of the delayed office items, thereby realizing office state self-adaptation of office item identification.

In some optional embodiments, the big data server adjusts the parameters of the office event detection model by maximizing the office performance progress detection results. In some alternative embodiments, the big data server adjusts the parameters of the first office event identification policy by minimizing the office execution progress detection result, since the error generated by the office execution progress detection result is also reflected in the first office event identification policy.

It should be noted that the embodiments of the present application are described only by taking the example of training the office event detection model at the same time when training the first office event recognition policy. In another embodiment, the office event detection model is a trained office event detection model, and the big data server does not need to train the office event detection model any more, then the big data server does not perform steps 208-210.

Further, the following is an implementation manner of determining an office execution progress provided by the embodiment of the present application, for example, the big data server inputs the real-time office transaction into the office transaction identification policy to obtain a first-state office transaction, and inputs the delayed office transaction into the office transaction identification policy to obtain a third-state office transaction. Wherein the big data server determines a first real-time office execution progress based on the determined node status weight feature map in case the real-time office event is determined to be an abnormal real-time office event, and the big data server does not need to utilize the node status weight feature map when determining a second real-time office execution progress in case the real-time office event is determined to be a normal real-time office event. And the big data server respectively inputs the first state office affairs and the third state office affairs into the office affair detection model, and determines an office execution progress detection result based on the detection result and the affair heat change value.

And step 211, the big data server performs office item scheduling processing based on the office item identification strategy obtained after training.

The office item identification strategy obtained after training has high accuracy and anti-interference performance, the big data server carries out office item scheduling processing on any office item based on the office item identification strategy, and differences and relevance among different office items can be fully considered in the office item scheduling process, so that the rationality of office item scheduling is ensured on the global level, normal execution of different office items is ensured as far as possible, and the situation that office items conflict with each other in the scheduling process is avoided.

In one possible implementation manner, the big data server performs office item identification on any office item based on the office item identification strategy obtained after training, obtains a state office item corresponding to any office item, and completes the office item identification task for any office item.

In step 212, the big data server determines the first real-time office event as a normal real-time office event in response to the first real-time office execution schedule being less than the target office execution schedule.

After the big data server determines the first real-time office execution progress, in response to that the first real-time office execution progress is smaller than the target office execution progress, the first real-time office item corresponding to the first real-time office execution progress is determined to be a normal real-time office item, and a subsequent second office item identification strategy can be trained based on the first real-time office item of the real-time office item determined to be normal. The target office execution progress is set by the big data server, and the embodiment of the application does not limit the progress.

In a possible implementation manner, in the same training turn, the big data server performs office item identification on a plurality of real-time office items to obtain corresponding real-time office execution schedules, and then the big data server selects a target number of real-time office execution schedules according to the priority order of task execution from the plurality of determined real-time office execution schedules, determines the real-time office items corresponding to the plurality of selected real-time office execution schedules as normal real-time office items, and determines the rest real-time office items as abnormal real-time office items.

Aiming at the condition that the real-time office matters corresponding to the real-time office matters in the training set are abnormal, the big data server corrects the abnormal states of the real-time office matters based on the real-time office execution progress, selects the normal real-time office matters corresponding to the normal states, improves the accuracy of the states, improves the signal-to-noise ratio of the office matters, and accordingly improves the accuracy of the office matter identification strategy obtained through training.

It should be noted that, in the embodiment of the present application, the determination of the office event is exemplified by executing the step 212 after the first office event identification policy is used, and in another embodiment, the big data server executes the step 202 to determine the first real-time office execution progress, and then executes the step 212 to determine the office event.

In step 213, the big data server determines a fifth real-time office execution schedule based on the real-time office event and the sixth real-time office event corresponding to the first real-time office event.

After the big data server determines the first real-time office event as a normal real-time office event based on the first office event identification strategy, the normal real-time office event can be applied to the training of the second office event identification strategy, and the fifth real-time office execution progress is determined based on the real-time office event and the sixth state office event corresponding to the first real-time office event. And the sixth state office event is obtained by re-identifying the office event of the first real-time office event based on the second office event identification strategy. The step 213 is the same as the process of determining the second real-time office execution progress in the step 204, and is not described in detail herein.

In step 214, the big data server determines a second delayed office execution progress based on the third status office transaction and the seventh status office transaction.

And the big data server applies the third state office affair obtained by carrying out office affair identification on the delayed office affair by the delayed office affair and the first office affair identification strategy to the training of the second office affair identification strategy, and determines a second delayed office execution progress based on the third state office affair and the seventh state office affair. And the seventh state office event is obtained by re-identifying the office event of the delayed office event based on the second office event identification strategy. The process of step 214 is the same as the process of determining the first delayed office execution progress in step 206, and is not repeated herein.

Step 215, the big data server trains a second office event recognition strategy based on the fifth time office execution schedule and the second delayed office execution schedule.

The process of step 215 is similar to the process of training the first office event recognition strategy in step 207, and is not repeated herein.

It should be noted that, in the embodiment of the present application, the second office event recognition strategy is trained based on the fifth office execution schedule and the second delayed office execution schedule. In another embodiment, the second office event identification policy also corresponds to an office event detection model, and after performing step 215, the method further comprises: the big data server carries out office item detection processing on the sixth state office item and the seventh state office item respectively based on an office item detection model corresponding to the second office item identification strategy to obtain a real-time detection result and a delay detection result, determines an office execution progress detection result based on the real-time detection result, the delay detection result and the seventh state office item, and trains the second office item identification strategy and the office item detection model corresponding to the second office item identification strategy based on the office execution progress detection result. The above process is similar to steps 208 to 210, and is not described herein again.

The embodiment of the present application will be described by taking as an example only the case where the first real-time office event is determined to be a normal real-time office event based on the first office event identification policy. In another embodiment, the big data server determines the first real-time office event as an anomalous real-time office event based on the first office event identification policy, then step 212-step 215 are replaced by the following step 216-step 219 as follows:

step 216, the big data server determines the first real-time office event as an abnormal real-time office event in response to the first real-time office execution schedule not being less than the target office execution schedule.

After the big data server determines the first real-time office execution progress, in response to that the first real-time office execution progress is not smaller than the target office execution progress, the first real-time office item corresponding to the first real-time office execution progress is determined to be an abnormal real-time office item, and a subsequent second office item identification strategy can be trained based on the first real-time office item of the real-time office item determined to be abnormal.

In one possible implementation, the big data server selects a plurality of first candidate real-time office events among the plurality of real-time office events based on the first office event recognition policy when training the first office event recognition policy, and selects a plurality of second candidate real-time office events among the plurality of real-time office events based on the second office event recognition policy when training the second office event recognition policy. The big data server determines the same real-time office event of the plurality of first candidate real-time office events and the plurality of second candidate real-time office events as an abnormal real-time office event.

Step 217, the big data server determines a sixth real-time office execution progress based on the real-time office affair, the first state office affair and the sixth state office affair corresponding to the first real-time office affair.

After the big data server determines the first real-time office item as the abnormal real-time office item based on the first office item identification strategy, the abnormal real-time office item can be applied to the training of the second office item identification strategy, and the sixth real-time office execution progress is determined based on the real-time office item, the first state office item and the sixth state office item corresponding to the first real-time office item. And the sixth state office event is obtained by re-identifying the office event of the first real-time office event based on the second office event identification strategy. The step 217 is the same as the process of determining the first real-time office execution progress in the step 202, and is not described herein again.

Step 218, the big data server determines a second delayed office execution schedule based on the third status office transaction and the seventh status office transaction.

Step 219, the big data server trains a second office item recognition strategy based on the sixth real-time office execution progress and the second delayed office execution progress.

The process from step 218 to step 219 is the same as the process from step 214 to step 215, and is not described in detail here.

In the embodiment of the present application, only the same real-time office event and delayed office event are used to train the first office event identification policy and the second office event identification policy, but in the process from the beginning of training to the completion of training, the big data server may train the first office event identification policy and the second office event identification policy by using multiple real-time office events and multiple delayed office events in the same training set.

In the embodiment of the application, aiming at abnormal state office matters in a training set and distribution difference information between real-time office matters and delayed office matters, an unsupervised anti-interference performance identification method based on an office state self-adaptive mechanism is provided, the abnormal real-time office matters with large loss in the real-time office matters are supervised by the same party to generate potential state office matters, thus the presence of an abnormal real-time office event is a global status office event and a predicted potential status office event, by learning and supervising the two office affair recognition strategies in an antagonistic training mode, a more reasonable training rule is designed, therefore, the problem that the abnormal office events are stored in the real-time office events and the problem of identification conversion from the real-time office events to the delayed office events are solved, and the accuracy and the anti-interference performance of the office event identification strategy can be improved.

With reference to the following example, in an embodiment of the present application, a big data server trains a first office event recognition policy and a second office event recognition policy by using a real-time office event and a delayed office event in the same training set, where the first office event recognition policy corresponds to a first office event detection model, the first office event detection model detects a status office event output by the first office event recognition policy, the second office event recognition policy corresponds to a second office event detection model, and the second office event detection model detects a status office event output by the second office event recognition policy.

The big data server inputs the real-time office matters into a first office matter identification strategy for processing, determines a real-time office execution progress, inputs the delayed office matters into the first office matter identification strategy for processing, determines a delayed office execution progress, inputs a result output by the first office matter identification strategy into a first office matter detection model for processing, determines an office execution progress detection result, and trains the first office matter identification strategy and the first office matter detection model based on the real-time office execution progress, the delayed office execution progress and the office execution progress detection result. Further, the big data server determines normal real-time office matters and abnormal real-time office matters in the real-time office matters based on the real-time office execution progress.

Similarly, the big data server respectively inputs the real-time office matters and the delayed office matters into the second office matter identification strategy for processing, determines the real-time office execution progress and the delayed office execution progress, inputs the results output by the second office matter identification strategy into the second office matter detection model for processing, determines the office execution progress detection result, and trains the second office matter identification strategy and the second office matter detection model based on the real-time office execution progress, the delayed office execution progress and the office execution progress detection result. Further, the big data server determines normal real-time office matters and abnormal real-time office matters in the real-time office matters based on the real-time office execution progress. Further, the big data server determines the intersection of the abnormal real-time office items and the abnormal real-time office items to obtain the abnormal real-time office items with high evaluation degree, re-determines the potential state office items of the abnormal real-time office items with high evaluation degree, and applies the potential state office items to the next training turn of the office item identification strategy.

According to the method provided by the embodiment of the application, when the real-time office matters and the delayed office matters are adopted to train the first office matter identification strategy, the identification result of the first office matter identification strategy is considered, the identification result of the second office matter identification strategy is also merged into the real-time office execution progress and the delayed office execution progress, so that the first office matter identification strategy learns the identification result of the second office matter identification strategy, the training process of the first office matter identification strategy can be supervised through the identification result of the second office matter identification strategy, and the first office matter identification strategy with higher accuracy and interference resistance can be trained.

Further, when the first real-time office event is determined to be an abnormal real-time office event, the real-time status office event corresponding to the abnormal real-time office event is not accurate enough, so that when the real-time office execution progress corresponding to the abnormal real-time office event is determined, the identification result of the second office event identification strategy for the abnormal real-time office event is also considered, and the first office event identification strategy is supervised by the second office event identification strategy, so that errors caused by the abnormal real-time office event in the training process can be reduced, and the anti-interference performance of the office event identification strategy for abnormal data can be improved.

Further, in order to ensure the accuracy of the generated potential office affairs, in the embodiment of the application, the first office affair recognition strategy and the second office affair recognition strategy are trained in a network parameter optimization manner, so that more accurate potential office affairs are continuously generated, and the potential office affairs are utilized to recognize the office affairs, so that the accuracy of the office affair recognition strategy is further improved.

Further, because the two office event identification strategies have different structures and learning capabilities, different types of errors introduced by abnormal office events can be filtered through the countertraining mode, so that the predicted potential state office event has stronger anti-interference performance. In the office item exchange process of the two office item identification strategies, the two office item identification strategies can supervise each other, filter abnormal office items in a training set and reduce training errors caused by the abnormal office items.

Furthermore, in the embodiment of the application, the item heat change value of the status office item output by the office item identification strategy is integrated into the detection loss, so that the counterstudy process of the office item identification strategy and the office item detection model is strengthened, the trained office item identification strategy can better learn the office status adaptive task capability, and the accuracy of the office item identification strategy is improved.

After the first office event identification strategy is trained, office event scheduling processing can be performed based on the first office event identification strategy. The following examples will describe the office event recognition process in detail.

The following embodiments of the application provide a big data office business processing method combining cloud computing. The execution subject of the embodiment of the application is a big data server, and the method comprises the following steps:

step 301, the big data server obtains a first office event identification policy.

The big data server obtains a first office affair identification strategy obtained after training, the first office affair identification strategy comprises a content verification network and a content identification network, the content verification network is used for carrying out content verification on real-time office affairs or delayed office affairs, and the content identification network is used for carrying out content identification on office affair feature contents.

Wherein, the training sample that adopts when training first office affairs recognition strategy includes: the real-time office affairs are respectively identified by the sample real-time office affairs, the sample real-time state office affairs corresponding to the sample real-time office affairs, the sample delayed office affairs, the first office affairs identification strategy and the second office affairs identification strategy to obtain the state office affairs.

Wherein, the training process of the first office event recognition strategy comprises the following steps: acquiring a sample real-time office item, a sample real-time state office item, a first state office item and a second state office item, wherein the first state office item is obtained by carrying out office item identification on the sample real-time office item based on a first office item identification strategy, and the second state office item is obtained by carrying out office item identification on the sample real-time office item based on a second office item identification strategy; determining a first real-time office execution progress based on the sample real-time office matters, the first state office matters and the second state office matters; acquiring a sample delayed office item, a third state office item and a fourth state office item, wherein the third state office item is obtained by carrying out office item identification on the sample delayed office item based on a first office item identification strategy, and the fourth state office item is obtained by carrying out office item identification on the sample delayed office item based on a second office item identification strategy; determining a first delayed office execution progress based on the third status office item and the fourth status office item; and training a first office item identification strategy based on the first real-time office execution progress and the first delayed office execution progress.

It should be noted that the sample real-time office event is similar to the first real-time office event in the foregoing embodiment, the sample real-time status office event is similar to the real-time status office event corresponding to the first real-time office event in the foregoing embodiment, the sample delayed office event is similar to the delayed office event in the foregoing embodiment, and the process of training the first office event identification policy is detailed in the foregoing embodiment, and is not described again here.

And step 302, the big data server performs content verification on the delayed office affairs based on the content verification network to obtain office affair feature contents corresponding to the delayed office affairs.

When any delayed office item needs to be identified, the big data server performs content verification on the delayed office item based on the content verification network in the first office item identification strategy to obtain office item feature content corresponding to the delayed office item, wherein the office item feature content is used for representing the content of the delayed office item.

Step 303, the big data server performs content identification on the office affair feature content based on the content identification network to obtain the delayed state office affair corresponding to the delayed office affair.

After the big data server obtains the office affair feature content, the content identification network identifies the office affair feature content based on the content in the first office affair identification strategy to obtain a delay state office affair corresponding to the delay office affair, and the class of the office interaction node in the delay office affair is marked in the delay state office affair, so that office affair identification processing of the delay office affair is completed.

In the method provided by the embodiment of the application, when the real-time office matters and the delayed office matters are adopted to train the first office matter identification strategy, the identification result of the first office matter identification strategy and the identification result of the second office matter identification strategy are considered, so that the first office matter identification strategy learns the identification result of the second office matter identification strategy, the training process of the first office matter identification strategy can be supervised by the identification result of the second office matter identification strategy, the first office matter identification strategy with higher accuracy and interference resistance can be trained, the difference and the relevance among different office matters can be fully considered in the office matter scheduling process, the rationality of office matter scheduling can be ensured on the global level, and the normal execution of different office matters can be ensured as much as possible, the situation that office matters conflict with each other in the scheduling process is avoided.

In some optional embodiments, after the office event scheduling processing is performed, the office event scheduling feedback uploaded by the office service end may be analyzed, so as to further dig out subsequent potential service requirements of the office service end, so that the office event scheduling can be preprocessed, thereby saving unnecessary time consumption during subsequent office event scheduling and improving the scheduling policy of subsequent office events. Based on this, on the basis of the above step 106, the method may further include the content described in step 107: acquiring office item scheduling feedback uploaded by a target office business terminal, and mining office business requirements on the office item scheduling feedback to obtain an office business requirement mining result; and carrying out office item scheduling preprocessing according to the office business requirement mining result.

In some optional embodiments, the "mining office business requirements for the office event scheduling feedback to obtain the office business requirement mining result" described in step 107 may be further implemented by the following steps 1071 to 1074.

Step 1071, obtaining an office action operation data track from the office item scheduling feedback, and obtaining a potential action operation data track according to the office action operation data track; wherein the office activity operation data track comprises uninterrupted sets of office activity operation data, and the potential activity operation data track comprises uninterrupted sets of potential activity operation data.

Step 1072, combining the office behavior operation data track, obtaining an office behavior content track through a first behavior content recognition network included in an office behavior mining model, and obtaining a potential behavior content track through a second behavior content recognition network included in the office behavior mining model, wherein the potential behavior content track includes a plurality of potential behavior contents, and the office behavior content track includes a plurality of office behavior contents.

And 1073, combining the office behavior content track and the potential behavior content track, and acquiring the potential office business requirements corresponding to the office behavior operation data track through an office business requirement identification layer included in the office behavior mining model.

In some optional embodiments, the step of "obtaining, by combining the office behavior content track and the potential behavior content track, the potential office business requirement corresponding to the office behavior operation data track through the office business requirement recognition layer included in the office behavior mining model" may be implemented by the following two implementation manners.

First embodiment

The step of obtaining the potential office business requirements corresponding to the office behavior operation data tracks through an office business requirement identification layer included in the office behavior mining model by combining the office behavior content tracks and the potential behavior content tracks includes: acquiring a plurality of first office behavior classification labels through a first behavior content screening layer included in the office behavior mining model by combining the office behavior content track, wherein each first office behavior classification label corresponds to one office behavior content; obtaining a plurality of second office behavior classification labels through a second behavior content screening layer included in the office behavior mining model in combination with the potential behavior content track, wherein each second office behavior classification label corresponds to one potential behavior content; performing label association processing on the plurality of first office behavior classification labels and the plurality of second office behavior classification labels to obtain a plurality of target office behavior classification labels, wherein each target office behavior classification label comprises a first office behavior classification label and a second office behavior classification label; obtaining global office behavior classification labels through a global behavior attention network included in the office behavior mining model by combining the plurality of target office behavior classification labels, wherein the global office behavior classification labels are determined according to the plurality of target office behavior classification labels and a plurality of office business heat values, and each target office behavior classification label corresponds to one office business heat value; and combining the global office behavior classification label, and acquiring the potential office business requirements corresponding to the office behavior operation data track through an office business requirement identification layer included in the office behavior mining model.

Second embodiment

The step of obtaining the potential office business requirements corresponding to the office behavior operation data tracks through an office business requirement identification layer included in the office behavior mining model by combining the office behavior content tracks and the potential behavior content tracks includes: acquiring a plurality of first office behavior classification labels through a first global business concern network included in the office behavior mining model by combining the office behavior content track, wherein each first office behavior classification label corresponds to one office behavior content; acquiring a plurality of second office behavior classification labels through a second global business concern network included in the office behavior mining model in combination with the potential behavior content track, wherein each second office behavior classification label corresponds to one potential behavior content; performing label association processing on the plurality of first office behavior classification labels and the plurality of second office behavior classification labels to obtain a plurality of target office behavior classification labels, wherein each target office behavior classification label comprises a first office behavior classification label and a second office behavior classification label; and combining the plurality of target office behavior classification labels, and acquiring potential office business requirements corresponding to the office behavior operation data tracks through the office business requirement identification layer included in the office behavior mining model.

It can be understood that, when the technical scheme is implemented, the office behavior classification labels can be obtained through the global behavior attention network or the global business attention network, so that the determination of the potential office business requirements is realized based on the behavior level or the business level, the potential office business requirements can be determined from different angles, and the accurate and complete determination of the office business requirement mining result can be conveniently carried out subsequently.

It is to be understood that the above two methods for determining potential office business needs can be used alternatively, and the embodiment is not limited herein.

And 1074, determining an office business requirement mining result of the office behavior operation data track according to the potential office business requirement.

It can be understood that, by implementing the steps 1071 to 1074, the potential office business requirements of the office business end can be mined from the office affair scheduling feedback based on the office behavior mining model, so that the office business requirement mining result of the office behavior operation data track is determined as accurately as possible based on the potential office business requirements, thereby saving unnecessary time consumption during subsequent office affair scheduling and improving the scheduling strategy of subsequent office affairs.

In some alternative embodiments, the "pre-processing office event scheduling according to the office business requirement mining result" described in step 107 can be implemented by:

acquiring business demand distribution information of the office business demand mining result and business demand matching items;

under the condition that the office business requirement mining result contains the real-time requirement category is determined according to the business requirement distribution information, determining the similarity between each business requirement matching item under the non-real-time requirement category of the office business requirement mining result and each business requirement matching item under the real-time requirement category of the office business requirement mining result according to the business requirement matching items under the real-time requirement category of the office business requirement mining results of a plurality of associated office business ends and item priorities, and sorting the business requirement matching items under the non-real-time requirement category of the office business requirement mining result, which are similar to the business requirement matching items under the real-time requirement category, into corresponding real-time requirement categories;

under the condition that the current non-real-time requirement category of the office business requirement mining result comprises a plurality of business requirement matching items, determining the similarity between the business requirement matching items under the current non-real-time requirement category of the office business requirement mining result according to the business requirement matching items under the real-time requirement category of the office business requirement mining result of a plurality of associated office business ends and item priorities thereof, and filtering the business requirement matching items under the current non-real-time requirement category according to the similarity between the business requirement matching items; setting service requirement heat for each type of service requirement matching items obtained by filtering according to service requirement matching items and item priorities under real-time requirement categories of office service requirement mining results of a plurality of associated office service ends, and sorting each type of service requirement matching items into the real-time requirement categories represented by the service requirement heat;

performing office item scheduling preprocessing through the business requirement matching items under the real-time requirement category; the office item scheduling preprocessing comprises the step of adjusting the execution priority of the business requirement matching items under the real-time requirement category.

It can be understood that through the office item scheduling preprocessing, the updating processing of the business requirement matching items under the real-time requirement category can be realized, so that the office item scheduling preprocessing is performed through the business requirement matching items under the real-time requirement category, unnecessary time consumption can be saved during subsequent office item scheduling, and the scheduling strategy of the subsequent office items can be improved.

On the basis, please refer to fig. 3, the present application further provides a block diagram of a big data office business processing apparatus 300 combined with cloud computing, where the apparatus includes the following functional modules.

A real-time office transaction obtaining module 310, configured to obtain a first real-time office transaction, a real-time office transaction corresponding to the first real-time office transaction, a first status office transaction, and a second status office transaction, where the first status office transaction is obtained by performing office transaction identification on the first real-time office transaction based on a first office transaction identification policy, and the second status office transaction is obtained by performing office transaction identification on the first real-time office transaction based on a second office transaction identification policy;

a real-time office progress determining module 320, configured to determine a first real-time office execution progress based on the real-time office details corresponding to the first real-time office details, the first state office details, and the second state office details;

a delayed office event acquiring module 330, configured to acquire a delayed office event, a third status office event, and a fourth status office event, where the third status office event is obtained by performing office event identification on the delayed office event based on the first office event identification policy, and the fourth status office event is obtained by performing office event identification on the delayed office event based on the second office event identification policy;

a delayed office progress determination module 340, configured to determine a first delayed office execution progress based on the third status office item and the fourth status office item;

and an office event scheduling processing module 350, configured to train the first office event recognition policy based on the first real-time office execution schedule and the first delayed office execution schedule, and perform office event scheduling processing based on the office event recognition policy obtained after the training.

On the basis, please refer to fig. 4, based on the same inventive concept, the present application further provides a big data office business processing system 400 combined with cloud computing, where the system includes a big data server 10 and an office business terminal 20 that communicate with each other;

wherein the big data server 10 is configured to: acquiring a first real-time office item, a real-time office item corresponding to the first real-time office item, a first state office item and a second state office item, wherein the first state office item is obtained by performing office item identification on the first real-time office item based on a first office item identification strategy, and the second state office item is obtained by performing office item identification on the first real-time office item based on a second office item identification strategy; determining a first real-time office execution progress based on the real-time office matters corresponding to the first real-time office matters, the first state office matters and the second state office matters; acquiring a delayed office item, a third state office item and a fourth state office item, wherein the third state office item is obtained by performing office item identification on the delayed office item based on the first office item identification strategy, and the fourth state office item is obtained by performing office item identification on the delayed office item based on the second office item identification strategy; determining a first delayed office execution progress based on the third status office transaction and the fourth status office transaction; and training the first office item identification strategy based on the first real-time office execution progress and the first delayed office execution progress, and performing office item scheduling processing based on the trained office item identification strategy.

Further, a readable storage medium is provided, on which a program is stored, which when executed by a processor implements the method described above.

It will be understood that the present application is not limited to the precise arrangements described above and shown in the drawings and that various modifications and changes may be made without departing from the scope thereof. The scope of the application is limited only by the appended claims.

Claims

1. A big data office business processing method combined with cloud computing is characterized by comprising the following steps:

2. The method of claim 1, wherein determining a first real-time office execution schedule based on the real-time office event, the first status office event, and the second status office event corresponding to the first real-time office event comprises:

in a case where the first real-time office event is determined to be an abnormal real-time office event based on the second office event identification policy, the first real-time office execution schedule is determined based on the real-time status office event, the first status office event, and the second status office event.

3. The method of claim 1, further comprising:

acquiring a second real-time office item, a real-time office item corresponding to the second real-time office item and a fifth state office item, wherein the fifth state office item is obtained by performing office item identification on the second real-time office item based on the first office item identification strategy;

determining a second real-time office execution progress based on the real-time office item corresponding to the second real-time office item and the fifth status office item, in a case where the second real-time office item is determined to be a normal real-time office item based on the second office item identification policy;

training the first office event recognition strategy based on the first real-time office execution progress, the second real-time office execution progress and the first delayed office execution progress.

4. The method of claim 1, wherein determining a first real-time office execution schedule based on the real-time office event, the first status office event, and the second status office event corresponding to the first real-time office event comprises:

determining a third real-time office execution schedule based on the real-time office event and the first status office event;

determining a fourth real-time office execution progress based on the second status office transaction and the first status office transaction;

and determining the first real-time office execution progress based on the third real-time office execution progress and the fourth real-time office execution progress.

5. The method of claim 4, wherein determining the first real-time office execution schedule based on the third real-time office execution schedule and the fourth real-time office execution schedule comprises:

acquiring network parameter optimization times corresponding to current network training;

and in response to the network parameter optimization times not being less than a first set time and not being more than a second set time, determining the first real-time office execution progress based on the third real-time office execution progress, the fourth real-time office execution progress, the network parameter optimization times, the first set time and the second set time.

6. The method of claim 1, wherein the third status office transaction includes a third status corresponding to each office interaction node in the delayed office transaction, wherein the fourth status office transaction includes a fourth status corresponding to each office interaction node in the delayed office transaction, and wherein determining a first delayed office execution schedule based on the third status office transaction and the fourth status office transaction includes:

determining a second office state comparison result based on a third state and a fourth state corresponding to each office interaction node, wherein the second office state comparison result represents difference information between the third state office item and the fourth state office item;

determining a local delay execution progress corresponding to any office interaction node based on the second office state comparison result, and a third state and a fourth state corresponding to any office interaction node;

determining the first delayed office execution progress based on the determined plurality of local delayed execution progresses.

7. The method of claim 1, further comprising:

performing office item detection processing on the first state office item based on an office item detection model to obtain a real-time detection result;

performing office item detection processing on the third state office item based on the office item detection model to obtain a delay detection result;

determining an office execution progress detection result based on the real-time detection result, the delay detection result and the third state office item;

and training the first office affair recognition strategy and the office affair detection model based on the office execution progress detection result.

8. The method of claim 1, wherein after determining a first real-time office execution schedule based on the real-time office event, the first status office event, and the second status office event corresponding to the first real-time office event, the method further comprises:

determining the first real-time office event as a normal real-time office event in response to the first real-time office execution progress being less than a target office execution progress;

or, in response to the first real-time office execution progress not being smaller than the target office execution progress, determining the first real-time office event as an abnormal real-time office event;

wherein, after determining the first real-time office event as an abnormal real-time office event in response to the first real-time office execution schedule not being less than the target office execution schedule, the method further comprises:

determining a sixth real-time office execution progress based on the real-time office affairs corresponding to the first real-time office affairs, the first state office affairs and a sixth state office affairs, wherein the sixth state office affairs are obtained by performing office affair identification on the first real-time office affairs based on the second office affair identification strategy;

determining a second delayed office execution progress based on the third state office transaction and a seventh state office transaction, wherein the seventh state office transaction is obtained by performing office transaction identification on the delayed office transaction based on the second office transaction identification policy;

training the second office event recognition strategy based on the sixth real-time office execution schedule and the second delayed office execution schedule.

9. The method of claim 1, further comprising:

acquiring office item scheduling feedback uploaded by a target office business terminal;

performing office business requirement mining on the office item scheduling feedback to obtain an office business requirement mining result;

carrying out office item scheduling preprocessing according to the office business requirement mining result;

the method for mining the office business requirements by the office item scheduling feedback to obtain the office business requirement mining result comprises the following steps:

acquiring an office behavior operation data track from the office item scheduling feedback, and acquiring a potential behavior operation data track according to the office behavior operation data track; the office behavior operation data track comprises multiple uninterrupted sets of office behavior operation data, and the potential behavior operation data track comprises multiple uninterrupted sets of potential behavior operation data;

in combination with the office behavior operation data track, acquiring an office behavior content track through a first behavior content identification network included in an office behavior mining model, and acquiring a potential behavior content track through a second behavior content identification network included in the office behavior mining model, wherein the potential behavior content track comprises a plurality of potential behavior contents, and the office behavior content track comprises a plurality of office behavior contents;

combining the office behavior content track and the potential behavior content track, and acquiring potential office business requirements corresponding to the office behavior operation data track through an office business requirement identification layer included in the office behavior mining model;

determining an office business requirement mining result of the office behavior operation data track according to the potential office business requirement;

the step of obtaining the potential office business requirements corresponding to the office behavior operation data tracks through an office business requirement identification layer included in the office behavior mining model by combining the office behavior content tracks and the potential behavior content tracks includes:

acquiring a plurality of first office behavior classification labels through a first behavior content screening layer included in the office behavior mining model by combining the office behavior content track, wherein each first office behavior classification label corresponds to one office behavior content;

obtaining a plurality of second office behavior classification labels through a second behavior content screening layer included in the office behavior mining model in combination with the potential behavior content track, wherein each second office behavior classification label corresponds to one potential behavior content;

performing label association processing on the plurality of first office behavior classification labels and the plurality of second office behavior classification labels to obtain a plurality of target office behavior classification labels, wherein each target office behavior classification label comprises a first office behavior classification label and a second office behavior classification label;

obtaining global office behavior classification labels through a global behavior attention network included in the office behavior mining model by combining the plurality of target office behavior classification labels, wherein the global office behavior classification labels are determined according to the plurality of target office behavior classification labels and a plurality of office business heat values, and each target office behavior classification label corresponds to one office business heat value;

and combining the global office behavior classification label, and acquiring the potential office business requirements corresponding to the office behavior operation data track through an office business requirement identification layer included in the office behavior mining model.

10. A big data server, comprising a processor and a memory; the processor is connected in communication with the memory, and the processor is configured to read the computer program from the memory and execute the computer program to implement the method of any one of claims 1 to 9.