CN115115093A - Object data processing method and device, electronic equipment and storage medium - Google Patents
Object data processing method and device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN115115093A CN115115093A CN202210555254.4A CN202210555254A CN115115093A CN 115115093 A CN115115093 A CN 115115093A CN 202210555254 A CN202210555254 A CN 202210555254A CN 115115093 A CN115115093 A CN 115115093A
- Authority
- CN
- China
- Prior art keywords
- intervention
- group
- result
- activation processing
- object group
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000003672 processing method Methods 0.000 title description 22
- 238000012545 processing Methods 0.000 claims abstract description 288
- 230000004913 activation Effects 0.000 claims abstract description 190
- 230000000694 effects Effects 0.000 claims abstract description 68
- 238000000034 method Methods 0.000 claims abstract description 52
- 230000002441 reversible effect Effects 0.000 claims abstract description 49
- 238000011282 treatment Methods 0.000 claims abstract description 43
- 230000004044 response Effects 0.000 claims description 41
- 230000015654 memory Effects 0.000 claims description 24
- 230000006870 function Effects 0.000 claims description 15
- 238000013145 classification model Methods 0.000 claims description 14
- 238000000605 extraction Methods 0.000 claims description 12
- 238000004590 computer program Methods 0.000 claims description 10
- 230000009467 reduction Effects 0.000 claims description 9
- 238000010801 machine learning Methods 0.000 claims description 8
- 238000006243 chemical reaction Methods 0.000 claims description 7
- 238000012549 training Methods 0.000 claims description 7
- 230000003213 activating effect Effects 0.000 claims description 2
- 238000001994 activation Methods 0.000 description 138
- 238000004364 calculation method Methods 0.000 description 16
- 238000005516 engineering process Methods 0.000 description 16
- 230000008569 process Effects 0.000 description 13
- 238000010586 diagram Methods 0.000 description 7
- 238000004891 communication Methods 0.000 description 5
- 238000013473 artificial intelligence Methods 0.000 description 4
- 230000006399 behavior Effects 0.000 description 4
- 230000009977 dual effect Effects 0.000 description 4
- 238000009434 installation Methods 0.000 description 4
- 239000000126 substance Substances 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 238000004422 calculation algorithm Methods 0.000 description 3
- 238000012512 characterization method Methods 0.000 description 3
- 230000001419 dependent effect Effects 0.000 description 3
- 238000001514 detection method Methods 0.000 description 3
- 230000014759 maintenance of location Effects 0.000 description 3
- 238000007726 management method Methods 0.000 description 3
- 230000001364 causal effect Effects 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 230000014509 gene expression Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000013515 script Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000033228 biological regulation Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000003066 decision tree Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Software Systems (AREA)
- General Physics & Mathematics (AREA)
- Operations Research (AREA)
- Data Mining & Analysis (AREA)
- Tourism & Hospitality (AREA)
- Game Theory and Decision Science (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Development Economics (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Quality & Reliability (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Entrepreneurship & Innovation (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The application provides a method and a device for processing object data, electronic equipment and a storage medium; the method comprises the following steps: performing event prediction processing on the object data of the intervention object group and the comparison object group to obtain a first prediction result and a second prediction result; performing activation processing on the basis of the first prediction result and the object data of the intervention object group to obtain a first activation processing result, and performing activation processing on the basis of the second prediction result and the object data of the comparison object group to obtain a second activation processing result; respectively carrying out reverse activation processing on the first activation processing result and the second activation processing result to obtain a first reverse activation processing result and a second reverse activation processing result; taking the difference value between the first and second reverse activation processing results as the average processing effect of the two object groups; and when the average treatment effect is larger than the average treatment effect threshold value, taking the similar objects of the intervention object group as the target objects subjected to the intervention treatment. By the method and the device, the accuracy of determining the target object subjected to the intervention processing can be improved.
Description
Technical Field
The present application relates to artificial intelligence technologies, and in particular, to a method and an apparatus for processing object data, an electronic device, and a storage medium.
Background
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.
Average Treatment Effect (ATE) is commonly used in causal inference studies to calculate causal parameters, and is defined without reference to a study design or estimation program. In both observational studies and experimental study design using random assignment methods, one can estimate the average treatment effect in a number of ways. The scheme with low temporal bias in the related art influences the accuracy of determining the average processing effect, and further influences the accuracy of determining the target object based on the average processing effect.
Disclosure of Invention
Embodiments of the present application provide a method and an apparatus for processing object data, an electronic device, a computer-readable storage medium, and a computer program product, which can reduce a deviation of an average processing effect and improve accuracy of determining a target object to be subjected to an intervention process.
The technical scheme of the embodiment of the application is realized as follows:
an embodiment of the present application provides a method for processing object data, including:
respectively carrying out event prediction processing on object data of an intervention object group and object data of a comparison object group to obtain a first prediction result of the intervention object group and a second prediction result of the comparison object group, wherein the event prediction processing is used for predicting a response result aiming at an event;
performing activation processing on the basis of the first prediction result and the object data of the intervention object group to obtain a first activation processing result of the intervention object group, and performing activation processing on the basis of the second prediction result and the object data of the control object group to obtain a second activation processing result of the control object group;
performing reverse activation processing on the first activation processing result and the second activation processing result to obtain a first reverse activation processing result of the intervention object group and a second reverse activation processing result of the control object group;
taking the difference value between the first reverse activation processing result of the intervention object group and the second reverse activation processing result of the control object group as the average processing effect between the intervention object group and the control object group;
when the average treatment effect is larger than an average treatment effect threshold value, determining similar objects of the intervention object group based on the average treatment effect, and regarding the similar objects as target objects subjected to intervention treatment.
An embodiment of the present application provides an apparatus for processing object data, including:
the result prediction module is configured to perform event prediction processing on object data of an intervention object group and object data of a comparison object group respectively to obtain a first prediction result of the intervention object group and a second prediction result of the comparison object group, wherein the event prediction processing is used for predicting a response result to an event;
an activation processing module configured to perform activation processing based on the first prediction result and the object data of the intervention object group to obtain a first activation processing result of the intervention object group, and perform activation processing based on the second prediction result and the object data of the control object group to obtain a second activation processing result of the control object group;
a reverse activation processing module configured to perform reverse activation processing on the first activation processing result and the second activation processing result to obtain a first reverse activation processing result of the intervention object group and a second reverse activation processing result of the control object group;
a comparison module configured to use a difference between a first reverse activation processing result of the intervention object group and a second reverse activation processing result of the control object group as an average processing effect between the intervention object group and the control object group;
and the matching module is configured to determine similar objects of the intervention object group based on the average treatment effect when the average treatment effect is larger than an average treatment effect threshold value, and take the similar objects as target objects subjected to intervention treatment.
An embodiment of the present application provides an electronic device, which includes:
a memory for storing executable instructions;
and the processor is used for realizing the object data processing method in the embodiment of the application when the executable instructions stored in the memory are executed.
The embodiment of the present application provides a computer-readable storage medium, which stores executable instructions, and when the executable instructions are executed by a processor, the method for processing object data according to the embodiment of the present application is implemented.
The embodiment of the present application provides a computer program product, which includes a computer program or instructions, and the computer program or instructions, when executed by a processor, implement the object data processing method according to the embodiment of the present application.
The embodiment of the application has the following beneficial effects:
the prediction results respectively corresponding to the intervention object group and the comparison object group are subjected to activation processing and inverse activation processing, so that the prediction results are converted from binary results into continuous results, the value range of the average processing effect is compressed through the activation processing, the deviation of double steady estimation is reduced, the accuracy of determining the target object is improved, and the effectiveness of applying the intervention processing is improved.
Drawings
Fig. 1 is an application mode schematic diagram of an object data processing method provided in an embodiment of the present application;
fig. 2 is a schematic structural diagram of an electronic device provided in an embodiment of the present application;
fig. 3A is a schematic flowchart of a method for processing object data according to an embodiment of the present application;
fig. 3B is a schematic flowchart of a method for processing object data according to an embodiment of the present application;
fig. 3C is a schematic flowchart of a method for processing object data according to an embodiment of the present application;
fig. 4 is a schematic diagram of distributed computing of a processing method of object data according to an embodiment of the present application;
fig. 5 is an alternative flow chart of a processing method of object data according to an embodiment of the present application;
fig. 6 is a schematic deviation scale diagram of a processing method of object data according to an embodiment of the present application.
Detailed Description
In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be described in further detail with reference to the attached drawings, the described embodiments should not be considered as limiting the present application, and all other embodiments obtained by a person of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.
In the following description, reference is made to "some embodiments" which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict.
In the following description, references to the terms "first \ second \ third" are only to distinguish similar objects and do not denote a particular order, but rather the terms "first \ second \ third" are used to interchange specific orders or sequences, where appropriate, so as to enable the embodiments of the application described herein to be practiced in other than the order shown or described herein.
It should be noted that, in the embodiments of the present application, the data related to the user information, the user feedback data, and the like, when the embodiments of the present application are applied to specific products or technologies, user permission or consent needs to be obtained, and the collection, use, and processing of the related data need to comply with the relevant laws and regulations and standards of the relevant countries and regions.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of the present application only and is not intended to be limiting of the application.
Before further detailed description of the embodiments of the present application, terms and expressions referred to in the embodiments of the present application will be described, and the terms and expressions referred to in the embodiments of the present application will be used for the following explanation.
1) The average result is the average of the response results of each subject to the event in the control group.
2) Average Treatment Effect (ATE): the method is used for characterizing the difference effect of different treatment means (intervention means) in random experiments, external intervention measures or medical experiments on the response result of an intervention object group and the average response result of a control object group. ATE was measured, i.e. the difference in the mean results for the events between the intervention and control groups was measured. In a random trial, the average treatment effect can be estimated from the samples by comparing the average results of the intervention and control groups.
3) And matching, and searching similar non-target users for each target user given the target users. For example: the user who purchases the commodity is the target user, the user who does not purchase the commodity is the non-target user, and similar users of the target user are matched in the user who does not purchase the commodity. And sending recommendation information for recommending commodities to the similar users, wherein the similar users may purchase commodities, and the recommendation effect of the recommendation information is improved by matching the similar users.
4) A trend Score (proportionality Score), which is used in embodiments of the present application to characterize the probability of an object being intervened, for example: the object is a user, and the probability that the user is pushed recommendation information can be determined based on the object characteristics.
5) And the confusion factor is a common reason for the change of the independent variable and the dependent variable. For example: when the independent variable is whether the coupon is obtained or not, the dependent variable is whether the coupon is purchased or not. The user's disposable resource limit has an effect on whether a coupon is obtained or not, and at the same time, the user's disposable resource limit has an effect on whether a purchase is obtained or not, and therefore, the user's disposable resource limit has an effect on both an independent variable (obtaining a coupon) and a dependent variable (purchase), and the user's disposable resource limit is a confounding factor between obtaining a coupon and the purchase.
6) Distributed computing is a computing method in which a large amount of engineering data to be computed is divided into small pieces and each piece is executed by a plurality of computers. After each computing unit in the distributed computing system outputs a corresponding operation result, all the results are combined in a unified mode to obtain a data conclusion. The computing units may be processors on different nodes, or different processors on the same node, or different cores in the same processor.
7) And activating the processing, namely performing mapping processing on the variable through a Sigmoid function. The Sigmoid function is often used as an activation function of a neural network to map a variable between 0 and 1, and the value range of the Sigmoid function is (0, 1), which can be used for two classifications.
8) And (5) performing inverse activation processing, and converting the binary results into continuous numerical values.
The embodiment of the application provides an object data processing method, an object data processing device, an electronic device, a computer readable storage medium and a computer program product, wherein a prediction result is converted from a binary result into a continuous result by performing activation processing and inverse activation processing on the prediction result, and the value range of an average processing effect is compressed by the activation processing, so that the deviation of double stable estimation is reduced, the accuracy of determining a target object is improved, and the effectiveness of intervention processing on the object is improved.
An exemplary application of the electronic device provided in the embodiments of the present application is described below, and the electronic device provided in the embodiments of the present application may be implemented as various types of user terminals such as a notebook computer, a tablet computer, a desktop computer, a set-top box, a mobile device (for example, a mobile phone, a portable music player, a personal digital assistant, a dedicated messaging device, a portable game device), a vehicle-mounted terminal, and the like, and may also be implemented as a server. The embodiment of the application can be applied to various scenes, including but not limited to cloud technology, artificial intelligence, intelligent traffic, driving assistance and the like. The embodiment of the present application may be implemented by a server, or implemented by a terminal device and the server in a cooperative manner, and an exemplary application when the electronic device is implemented as the server will be described below.
Referring to fig. 1, fig. 1 is a schematic diagram of an application mode of processing object data provided in an embodiment of the present application; by way of example, reference is made to a server comprising: the recognition server 201, the recommendation server 202, the network 300 and the terminal device 401. The identification server 201 and the recommendation server 202 communicate with each other through the network 300 or communicate with each other in other ways, the terminal device 401 is connected to the identification server 201 through the network 300, and the network 300 may be a wide area network or a local area network, or a combination of the two.
For example, the control group is simply referred to as a control group, the intervention group is simply referred to as an intervention group, and the recognition server 201 acquires object data of the control group and object data of the intervention group. The recognition server 201 performs event prediction processing on the basis of the object data of the intervention object group and the object data of the comparison object group, respectively, to obtain a first prediction result of the intervention object group and a second prediction result of the comparison object group. The prediction results of each object group are respectively subjected to activation processing and inverse activation processing, the difference between the processing results of the intervention group and the comparison group is used as the average processing effect of the two, the average processing effect of the intervention group and the comparison group is sent to the recommendation server 202 through the network 300, the recommendation server 202 can determine similar objects corresponding to the objects in the intervention group based on the average processing effect, and the similar objects are used as target objects to send recommendation results (implement intervention behaviors).
In some embodiments, the intervening act is sending the commercial advertisement to a terminal device held by the subject, the control group is comprised of users who did not receive the advertisement, and the intervening group is comprised of users who received the advertisement. The event result is whether the user purchased the goods corresponding to the advertisement. The object data of the contrast group and the object data of the intervention group are identified by the server 201, an average processing effect between the intervention group and the object group is obtained based on object data analysis, the average processing effect is used for representing a difference effect of sent advertisements and unsent advertisements on an event of commodity purchase, when the average processing effect is larger than an average processing effect threshold value, the recommendation server 202 determines that association exists between the sent advertisements and the purchased commodities, determines similar users of the intervention group, sends the advertisements to the similar users, and further improves advertisement recommendation effect.
In some embodiments, the recognition server 201 and the recommendation server 202 may also be implemented as a unified server.
The embodiment of the application can be realized by a block chain technology, the abnormal account obtained by the object data processing method of the embodiment of the application can be used as a detection result, the detection result is uploaded to a block chain to be stored, and the reliability of the detection result is ensured by a consensus algorithm. The blockchain is a novel application mode of computer technologies such as distributed data storage, point-to-point transmission, a consensus mechanism and an encryption algorithm. A block chain (Blockchain), which is essentially a decentralized database, is a series of data blocks associated by using a cryptographic method, and each data block contains information of a batch of network transactions, so as to verify the validity (anti-counterfeiting) of the information and generate a next block. The blockchain may include a blockchain underlying platform, a platform product services layer, and an application services layer.
The embodiment of the application can be realized by a Database technology, wherein a Database (Database) can be regarded as a place where an electronic file is stored in an electronic file cabinet in short, and a user can add, query, update, delete and the like to data in the file. A "database" is a collection of data that is stored together in a manner that can be shared by multiple users, has as little redundancy as possible, and is independent of the application.
A Database Management System (DBMS) is a computer software System designed for managing a Database, and generally has basic functions such as storage, interception, security assurance, and backup. The database management system may classify the database according to the database model it supports, such as relational, XML (Extensible Markup Language); or classified according to the type of computer supported, e.g., server cluster, mobile phone; or classified according to the Query Language used, such as Structured Query Language (SQL), XQuery; or by performance impulse emphasis, e.g., maximum size, maximum operating speed; or other classification schemes. Regardless of the manner of classification used, some DBMSs are capable of supporting multiple query languages across categories, for example, simultaneously.
In some embodiments, the server may be an independent physical server, may also be a server cluster or a distributed system formed by a plurality of physical servers, and may also be a cloud server providing 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, and a big data and artificial intelligence platform. The terminal device may be, but is not limited to, a smart phone, a tablet computer, a notebook computer, a desktop computer, a smart speaker, a smart watch, and the like. The terminal device and the server may be directly or indirectly connected through wired or wireless communication, and the embodiment of the present application is not limited thereto.
The embodiment of the application can also be realized through a Cloud Technology, and a Cloud Technology (Cloud Technology) can form a resource pool based on a general term of a network Technology, an information Technology, an integration Technology, a management platform Technology, an application Technology and the like applied in a Cloud computing business model, and can be used as required, so that the Cloud Technology is flexible and convenient. Cloud computing technology will become an important support. Background services of the technical network system require a large amount of computing and storage resources, such as video websites, picture-like websites and more web portals. With the high development and application of the internet industry and the promotion of the requirements of search service, social network, mobile commerce, open collaboration and the like, each article may have a hash code identification mark, the hash code identification mark needs to be transmitted to a background system for logic processing, data at different levels can be processed separately, various industrial data need strong system background support, and the hash code identification mark can be realized only through cloud computing.
Referring to fig. 2, fig. 2 is a schematic structural diagram of an electronic device provided in an embodiment of the present application, including: at least one processor 410, memory 450, at least one network interface 420. The various components in electronic device 400 are coupled together by a bus system 440. It is understood that the bus system 440 is used to enable communications among the components. The bus system 440 includes a power bus, a control bus, and a status signal bus in addition to a data bus. For clarity of illustration, however, the various buses are labeled as bus system 440 in fig. 2.
The Processor 410 may be an integrated circuit chip having Signal processing capabilities, such as a general purpose Processor, a Digital Signal Processor (DSP), or other programmable logic device, discrete gate or transistor logic device, discrete hardware components, or the like, wherein the general purpose Processor may be a microprocessor or any conventional Processor, or the like.
The memory 450 may be removable, non-removable, or a combination thereof. Exemplary hardware devices include solid state memory, hard disk drives, optical disk drives, and the like. Memory 450 optionally includes one or more storage devices physically located remote from processor 410.
The memory 450 includes either volatile memory or nonvolatile memory, and may include both volatile and nonvolatile memory. The nonvolatile memory may be a Read Only Memory (ROM), and the volatile memory may be a Random Access Memory (RAM). The memory 450 described in embodiments herein is intended to comprise any suitable type of memory.
In some embodiments, memory 450 is capable of storing data, examples of which include programs, modules, and data structures, or a subset or superset thereof, to support various operations, as exemplified below.
The operating system 451, which includes system programs for handling various basic system services and performing hardware-related tasks, such as a framework layer, a core library layer, a driver layer, etc., is used for implementing various basic services and for processing hardware-based tasks.
A network communication module 452 for communicating to other computing devices via one or more (wired or wireless) network interfaces 420, exemplary network interfaces 420 including: bluetooth, wireless compatibility authentication (WiFi), and Universal Serial Bus (USB), among others.
In some embodiments, the processing device of the object data provided in this embodiment may be implemented in software, and fig. 2 illustrates the processing device 455 of the object data stored in the memory 450, which may be software in the form of programs and plug-ins, and includes the following software modules: a result prediction module 4551, an activation processing module 4552, a reverse activation processing module 4553, a comparison module 4554, and a matching module 4555, which are logical and thus can be arbitrarily combined or further split depending on the functions implemented. The functions of the respective modules will be explained below.
The object data processing method provided by the embodiment of the present application will be described in conjunction with exemplary applications and implementations of the terminal provided by the embodiment of the present application.
Referring to fig. 3A, fig. 3A is a schematic flowchart of a method for processing object data according to an embodiment of the present application, and will be described with reference to the steps shown in fig. 3A.
In step 301, event prediction processing is performed on the intervention target data of the intervention target group and the control target data of the control target group, respectively, to obtain a first prediction result of the intervention target group and a second prediction result of the control target group.
Here, the event prediction processing is for predicting a response result to an event.
For example, the event may be an event that occurs in an environment where the intervention object and the control object are located together, the intervention object being an object to which the intervention process is performed, and the control object being an object to which the intervention process is not performed. For example: the event is that the intervention object and the control object are both users using the same shopping platform, and the result of the event is that the user purchases the commodity or the user does not purchase the commodity. The intervention process may be to recommend an advertisement for article a to the subject, the intervention subject being the subject to which the advertisement for article a is pushed, and the control subject being the subject to which the advertisement for article a is not pushed. As a result, some of the intervention subjects in the intervention subject group and some of the control subjects in the control subject group purchased the article a, and another part of the intervention subjects and another part of the control subjects did not purchase the article a.
In some embodiments, step 301 may be implemented by: performing feature extraction on the intervention object data of the intervention object group to obtain the intervention object features of the intervention object group, and performing feature extraction on the comparison object data of the comparison object group to obtain the comparison object features of the comparison object group; and calling the result prediction model based on the characteristics of the intervention object to obtain a first prediction result of the intervention object group, and calling the result prediction model based on the characteristics of the comparison object to obtain a second prediction result of the comparison object group.
Example, the intervention object data includes: the actual response result of the intervention object, the attribute parameters of the intervention object and the intervention data; the control object data includes: comparing the actual response result of the object and the attribute parameter of the object. The outcome prediction model includes: the system comprises a first result prediction model used for extracting the characteristics of the intervention object and a second result prediction model used for extracting the characteristics of the control object.
For example, the feature extraction of the intervention object data of the intervention object group to obtain the intervention object features of the intervention object group may be implemented as follows: and performing conversion processing on the intervention object data of the intervention object group to obtain an object feature vector of the intervention object group, and performing dimension reduction processing on the object feature vector of the intervention object group to obtain the intervention object feature of the intervention object group.
For example, the feature extraction of the comparison object data of the comparison object group to obtain the comparison object features of the comparison object group may be implemented as follows: and performing conversion processing on the comparison object data of the comparison object group to obtain an object feature vector of the comparison object group, and performing dimension reduction processing on the object feature vector of the comparison object group to obtain the comparison object feature of the comparison object group.
As an example, the intervention object data and the comparison object data are data that cannot be read directly by a machine, and the intervention object data and the comparison object data may be converted into a feature vector (object feature vector) in a digital form that can be read by a computer through a conversion process. The dimension of the object feature vector is high, and the object feature vector can be converted into the object feature in the form of the unique heat vector through dimension reduction processing, so that subsequent calculation and processing are facilitated.
Illustratively, the preset logic classification model is trained based on the actual response result of the intervention object, the attribute parameter of the intervention object and the intervention data to obtain a first result prediction model, and the preset logic classification model is trained based on the actual response result of the comparison object and the attribute parameter of the comparison object to obtain a second result prediction model.
For example, continuing with the above example, the intervention object property parameter or the control object property parameter may include at least part of: the shopping frequency of the object on the shopping platform in a preset number of days (for example, 30 days), the browsing duration of the object on the shopping platform in the preset number of days, the bought-back goods of the object on the shopping platform, and the like. The intervention data may be the number of interventions applied to the intervention subject, the specific behavior of the interventions, the duration of the intervention behaviors, etc. The actual response result of the subject is whether the subject actually purchased the good. The prediction result is analyzed based on information such as an attribute parameter of the object, and there is a possibility that the prediction result deviates from the actual result to a certain extent.
In step 302, an activation process is performed based on the first prediction result and the intervention target data of the intervention target group to obtain a first activation process result of the intervention target group, and an activation process is performed based on the second prediction result and the control target data of the control target group to obtain a second activation process result of the control target group.
In some embodiments, referring to fig. 3B, step 302 may also be implemented through steps 3021 to 3025, which are described in detail below.
In step 3021, intervention probabilities corresponding to the intervention subject group and the control subject group are obtained.
In some embodiments, step 3021 may be implemented by: and calling the machine learning classification model to carry out intervention probability prediction processing on the basis of the intervention object attribute parameters of the intervention object group to obtain the intervention probability of each intervention object in the intervention object group, and calling the machine learning classification model to carry out intervention probability prediction processing on the basis of the comparison object attribute parameters of the comparison object group to obtain the intervention probability of each comparison object in the comparison object group.
For example, the intervention probability is the probability of whether an object will be handled by an intervention, such as: the intervention processing is to carry out advertisement pushing on the user, and the probability that the user views the advertisement corresponding to the intervention processing in the process of using the shopping platform is the intervention probability. The intervention probability ranges between 0 and 1.
In step 3022, a total activation processing outcome for the group of intervention subjects is determined based on the first prediction outcome, the intervention probability for each intervention subject in the group of intervention subjects, the actual response outcome for the intervention subject, and the intervention data.
In some embodiments, step 3022 may be implemented by: for each intervention object in the group of intervention objects, performing the following: determining a first intermediate parameter based on the reciprocal of the first prediction result, and determining a second intermediate parameter based on the reciprocal of the actual response result of the intervention object; taking the ratio of the intervention data to the intervention probability of the intervention object as a third intermediate parameter; acquiring a fourth difference value of the second intermediate parameter and the first intermediate parameter, multiplying the fourth difference value and the third intermediate parameter, adding the product obtained by multiplication and the first intermediate parameter, and taking the sum obtained by addition as an object activation processing result of an intervention object; and adding the object activation processing results of each intervention object to obtain the total activation processing result of the intervention object group.
For example, the result of the activation process for the intervention object group is calculated by equation (1), as follows:
wherein the content of the first and second substances,
is the result of the first activation process, T i Is intervention data, Y i Is the result of the actual response that is,
is the intervention probability for the intervention object i. N is the number of objects of the intervening object,
the prediction result corresponding to the ith intervention object in the intervention object group is obtained.
In step 3023, the total activation processing result of the intervention object group is divided by the number of intervention objects in the intervention object group to obtain a first activation processing result of the intervention object group.
In step 3024, a total activation treatment outcome for the group of control subjects is determined based on the second prediction outcome, the intervention probability for each control subject in the group of control subjects, and the actual response outcome for the control subject.
In some embodiments, step 3024 may be implemented by: the following treatments were performed for each control subject in the control subject group: determining a fourth intermediate parameter based on the inverse of the second predicted result and a fifth intermediate parameter based on the inverse of the actual response result; acquiring a first difference value between the intervention data of the intervention object corresponding to the probability 1 comparison object, taking the first difference value as the comparison object intervention data of the comparison object, acquiring a second difference value between the intervention probabilities of the probability 1 comparison object, and taking the ratio of the first difference value to the second difference value as a sixth intermediate parameter; acquiring a third difference value between the fifth intermediate parameter and the fourth intermediate parameter, multiplying the third difference value and the sixth intermediate parameter, adding the product obtained by multiplication and the fourth intermediate parameter, and taking the sum obtained by addition as an object activation processing result of a comparison object; and adding the object activation processing results of each comparison object to obtain the total activation processing result of the comparison object group.
For example, equation (2) is calculated against the activation processing result of the object group, as follows:
wherein the content of the first and second substances,
is the second activation process result, (1-T) i ) Is the first difference, the characterization is not intervened, Y i As a result of which it is possible to,
is the second difference between the probability 1 versus the intervention probability of the subject. N is the number of control objects,
is the second prediction of the ith control subject in the control subject group.
In step 3025, the total activation processing result of the control object group is divided by the number of control objects in the control object group to obtain a second activation processing result of the control object group.
For example, the activation processing results corresponding to the two object groups are obtained by processing according to the above formula (1) and formula (2).
In step 303, reverse activation processing is performed on the first activation processing result and the second activation processing result to obtain a first reverse activation processing result of the intervention object group and a second reverse activation processing result of the control object group.
In some embodiments, the first activation processing result and the second activation processing result are binary results; a binary result is also a result characterized by either a 1 or a 0. Step 303 may be implemented by: and taking the negative number of the second activation processing result as the independent variable of the activation function to carry out activation processing, so as to obtain a second reverse activation processing result with continuous value range.
For example, the activation function is also a sigmoid function, and further, the result of the first reverse activation process of the intervention group of the reverse activation process is characterized as
The second reverse activation treatment result of the reverse activation treated control group is characterized in that
In step 304, the difference between the first inverse activation processing result of the intervention object group and the second inverse activation processing result of the control object group is used as the average processing effect between the intervention object group and the control object group.
Illustratively, the first reverse activation process is concludedSubtracting the result from the second inverse activation result to obtain an average processing effect delta DR,B The calculation formula (3) is as follows:
in step 305, when the average processing effect is greater than the average processing effect threshold, similar objects of the intervention object group are determined based on the average processing effect, and the similar objects are used as target objects to be subjected to the intervention processing.
In some embodiments, referring to fig. 3C, step 305 may also be implemented by steps 3051 through 3054, detailed below.
In step 3051, an intervention probability of each intervention object in the set of intervention objects and an intervention probability corresponding to each sample object in the set of sample objects are obtained.
Here, each sample object in the sample object group is an object to which no intervention processing is performed.
For example, the control object of the control object group may be an object that has not been subjected to the intervention treatment. The source of the sample object in the sample object group may be a new object crawled from the network or a control object in the control object group.
In step 3052, a trend score is determined for each intervention object based on the intervention probability for each intervention object, and a trend score is determined for each sample object based on the intervention probability for each sample object.
In some embodiments, step 3052 can be implemented by: multiplying the intervention probability of each intervention object by the intervention parameters to obtain a trend value of each intervention object, rounding the trend value of each intervention object to obtain a trend score of each intervention object, multiplying the intervention probability of each sample object by the intervention parameters to obtain a trend value of each sample object, rounding the trend value of each sample object to obtain a trend score of each sample object.
For example, the intervention parameter may be 100. The following examples are given, assuming: the intervention probability of the intervention object is 0.31415, the trend value obtained by multiplying the intervention parameter by the intervention probability is 31.415, and the trend value is rounded up to obtain a trend score of 31. The trend scores of the sample objects may be obtained in the same manner and are not described in detail herein.
In step 3053, for each intervention object, the following is performed: and performing matching processing in the sample object group based on the trend score of the intervention object to obtain a sample object with the same trend score as the intervention object, and taking the sample object with the same trend score as the similar object of the intervention object.
For example, the trend scores are similar, which indicates that the object data corresponding to the objects are similar, and further, the intervention probabilities predicted based on the object data are similar, the trend scores are similar. In the embodiment of the application, the similarity between the objects is represented through the trend scores, and the computing resources are saved.
In step 3054, each similar object in the sample object group is taken as a target object to be subjected to the intervention process.
For example, it is assumed that the intervention is to push recommendation information to the user, the recommendation information may be an advertisement, and the object data processing method according to the embodiment of the present application may determine that there is an association between the recommendation of the advertisement to the object and the result of the object purchasing a product. The intervention object is an object pushed with the advertisement of the product A, the trend score of the intervention object is 59, the sample object with the trend score of 59 is a similar object of the intervention object, the similar object is used as a target object, the advertisement of the product A is pushed to the target object, the target object has a large possibility of purchasing the product A after watching the advertisement or remembering the product A, and then the advertisement recommendation is carried out through the method, so that the purchase rate of the product recommended by the advertisement is improved, and the advertisement recommendation effect is improved.
According to the method and the device, the prediction results corresponding to the intervention object group and the comparison object group are subjected to activation processing and inverse activation processing, so that the prediction results are converted from binary results into continuous results, the value range of the average processing effect is compressed through the activation processing, the deviation of double steady estimation is reduced, the accuracy of determining the target object is improved, and the effectiveness of applying the intervention processing is improved.
Next, an exemplary application of the embodiment of the present application in a practical application scenario will be described.
The object data processing method provided by the embodiment of the application can be applied to the following application scenes:
1. cause and effect inference analysis of backflow activity: the method comprises the following steps that users using the application program gradually lose, the lost users are possible to reflow (namely, the application program is reused) after being pushed with the short messages, the intervention group is users who are pushed with the short messages, the comparison group is users who are not pushed with the short messages, and whether the users reflow is judged as a result. Through the evaluation of the reflow activity recommended by different short messages, the optimal reflow activity can be found, limited push resources are allocated to the efficient reflow activity, and the reflow rate can be improved, for example: 3 to 5 percent of the total weight of the product.
2. The face recognition verifies the influence of game play on the user: in the game-play process, if the online time of the user is longer than the preset use time of the minor, face recognition verification may be required to determine whether the user participating in the game is an adult or a minor. In this scenario, the intervention is face recognition verification, the intervention group is users who have performed face recognition, the comparison group is users who have not performed face recognition, and the result is whether the users have performed game match after being subjected to face recognition. The method for processing object data provided by the embodiment of the application can obtain the influence of face recognition and game on the opening rate of the game, for example: face recognition reduces the opening rate by 1% -2%.
3. User report rate and retention analysis: after the account number of the user is reported on a certain platform/game, the intervention group comprises reported users, the comparison group comprises users which are not reported, the event result is whether the user reserves the platform or the game, and the influence of the reporting behavior on the platform/game reserved by the user can be obtained through calculation by the object data processing method provided by the embodiment of the application.
4. And (3) analyzing the recommendation effect: the method comprises the steps that a batch of users purchasing game props are provided, an intervention group is users purchasing the game props, a comparison group is users not purchasing the game props, and whether the subsequent launched game props in a game are purchased or not is judged as a result.
5. Playing analysis of the game: the intervention is that the user participates in a certain play in the game, the intervention group is composed of users who participate in the play, the contrast group is composed of users who do not participate in the play, and the result is whether the user retains the game. The relationship between the play and the retention rate can be obtained through the processing method of the object data provided by the embodiment of the application, for example: retention is increased by 10% with a certain play.
6. Analyzing the game content: after a batch of exposed game advertisements are given, the game installation amount of the users is increased, the intervention group is users who are pushed with the advertisements, the comparison group is users who are not pushed with the advertisements, and whether a game application program is installed or not is judged as a result, the influence of various game advertisement contents on the game installation rate is obtained through the object data processing method provided by the embodiment of the application, so that the limited advertisement exposure amount is distributed to the most efficient advertisement contents, and the installation rate is improved, for example: the installation rate is improved by 1-2%.
Referring to fig. 5, fig. 5 is an alternative schematic flowchart of a processing method of object data according to an embodiment of the present application.
In step 501, the intervention probability of the subject is predicted.
For example, the intervention probability refers to the possibility of an object being intervened, and the purpose of predicting the intervention probability is to convert differences of different dimensions of the object into one-dimensional differences so as to facilitate comparison and finding similar objects, which is specifically performed by: training a machine learning classification model GBDT (gradient Boosting Decision Tree) based on the object features X and the intervention t to obtain a model for predicting the intervention probability of the object, learning the relation between X and t, and predicting the (different types of intervention) intervention probabilities of different objects based on the trained machine learning classification model, wherein the intervention probability is a decimal number between 0 and 1.
For example, a trend Score (PS) may be calculated based on the intervention probability, with the formula: PS ≈ intervention probability 100, i.e., the intervention probability is multiplied by 100 and the resulting value is rounded, converting the intervention probability into an integer between 0 and 100. Each subject will have a PS (trend score) between 0 and 100, with a subject's PS towards 100 indicating that the subject is more trending to intervene. The trend scores can be used as the characteristics of the objects, and objects with similar or same trend scores are used as similar objects.
In some embodiments, step 501 may be performed before step 504, after steps 502 and 503.
In step 502, a result prediction is performed based on the object data of the intervention group to obtain a prediction result of the intervention group.
Illustratively, step 502 may be implemented by: training a result prediction model corresponding to the intervention group on the basis of the intervention group data
And calling the trained model to perform event prediction processing on the object data of the intervention group to obtain a prediction result (a first prediction result) of the intervention group.
By way of example, feature engineering is established using the object features Xi and the results Yi of the intervention group. The process of establishing feature engineering is the process of converting raw data into features characterizing a latent model of the data. The characteristic engineering is established in the following way: and performing feature extraction by using the object features of each object in the intervention group and the intervened result of each object as feature data to obtain object features (characterized by digital vector features) which can be learned by a machine, and performing dimension reduction processing on the object features of each object to obtain the unique heat vector features of each object in the intervention group. And training a result prediction model corresponding to the intervention group based on the unique heat vector characteristics of each intervention object, and performing result prediction on data of the intervention object to be predicted based on the trained result prediction model to obtain a prediction result of each object. The result prediction model may be trained using pyspark (application program interface for large-scale data processing).
In step 503, the result prediction is performed based on the target data of the control group, and the prediction result of the control group is obtained.
Illustratively, step 503 may be implemented by: training corresponding result prediction model of control group based on control group data
And calling the trained model to perform event prediction processing on the object data of the control group to obtain a prediction result (second prediction result) of the control group.
Illustratively, feature engineering is established using object features Xi and results Yi of the control group. The process of establishing feature engineering is the process of converting raw data into features characterizing a latent model of the data. The characteristic engineering is established in the following way: and performing feature extraction by using the object features of each object in the comparison group and the compared result of each object as feature data to obtain object features (characterized by digital vector features) which can be learned by a machine, and performing dimension reduction processing on the object features of each object to obtain the unique heat vector features of each object in the comparison group. And training a result prediction model corresponding to the comparison group based on the unique heat vector characteristics of each comparison object, and performing result prediction on data of the comparison object to be predicted based on the trained result prediction model to obtain a prediction result of each object. The result prediction model may be trained using pyspark (application program interface for large-scale data processing).
Referring to fig. 4, fig. 4 is a distributed computing representation of a processing method of object data provided by an embodiment of the present applicationAn intent; each box corresponding to each calculation formula in fig. 4 represents a calculation unit (calculation unit 401, calculation unit 402, calculation unit 403, calculation unit 404, calculation unit 405, calculation unit 406, and calculation unit 407), and the calculation processes of each calculation unit do not interfere with each other. Wherein the content of the first and second substances,
the predicted outcome of the control group was characterized,
the prediction results of the intervention group are characterized,
the difference between the predicted result and the actual result of the control group is characterized,
characterizing the difference between the predicted result and the actual result of the intervention group,
the relationship between the control object and the difference between the predicted results,
the relationship between the intervention objects and the difference between the predicted outcomes between the characterizations. The object data input by distributed computation, (X, t, y), wherein X represents the characteristics of the intervention object or the comparison object (such as the login times of the last 30 days, the login times of the last 7 days, the variety of the game installed in the last 30 days, the number of the games installed in the last 30 days and the like aiming at the game); t represents: whether the subject is intervened (e.g., whether to recommend an advertisement to the subject, the number of times the advertisement is recommended), y represents the result (e.g., whether the user has reflowed); and (3) outputting distributed computation: a result estimation table comprising the results of the matched control group of control subjects and intervention subjects of the intervention group,the result estimation table has the following fields (whether intervention, result). By taking 'whether to intervene' as a main key, the 'result' can be directly inquired.
In step 504, the activation process results for the intervention group are determined.
The result of the activation process for the intervention group is calculated by equation (1) as follows:
wherein, the first and the second end of the pipe are connected with each other,
is the result of activation processing, T i Is an intervention, Y i As a result of which it is possible to,
is the probability of intervention. N is the number of samples and,
is the predicted result of the intervention group. With continued reference to FIG. 4, equation (1) corresponds to the equation in FIG. 4 based on
A branch of the prediction result is taken.
In step 505, the activation treatment results for the control group are determined.
Equation (2) was calculated for the results of the activation treatment of the control group as follows:
wherein the content of the first and second substances,
is the result of activation processing, (1-T) i ) Characterization not intervened, Y i As a result of which it is possible to,
is the intervention probability obtained in step 501. N is the number of samples and,
is the predicted result of the control group. With continued reference to FIG. 4, equation (2) corresponds to the equation in FIG. 4 based on
A branch of the prediction result is taken.
In step 506, the activation treatment results of the intervention group and the control group are subjected to reverse activation treatment.
For example, the activation process results of the intervention group of the inverse activation process are characterized by
The activation treatment results of the reverse activation treated control group were characterized as
In step 507, the average treatment effect between the intervention group and the control group is determined.
Exemplary, average treatment Effect Δ DR,B The calculation formula (3) of (c) is as follows:
in some embodiments, after step 508, the influence of the intervention on the result may be further determined based on the average processing effect, and if the influence degree is greater than the preset influence degree, objects similar to the objects of the intervention group may be determined by using the trend score obtained in step 1, and then the objects may be selected for intervention to obtain a corresponding result.
Assuming that the object is a user, the intervention is advertisement recommendation, and the result is purchasing a commodity, based on the average processing effect obtained in steps 501 to 507 in this embodiment, it may be determined that the advertisement recommendation has a positive influence on commodity purchase (that is, the user is promoted to purchase a commodity), based on the trend score, the similar users of the users in the intervention group are determined, users similar to the users in the intervention group are obtained to perform advertisement recommendation, and then the similar users are obtained to perform advertisement recommendation, so that the advertisement recommendation effect may be improved.
Compared with the prior art, the effect of the processing method of the object data provided by the embodiment of the present application is embodied in the experimental data, and the processing method of the object data provided by the present application is referred to as Binary dual robust estimation hereinafter. In the context of a Binary result with a Hidden confusion factor (Hidden-provider), the mean bias of Binary dual robust estimation is lower, which is 42.16% lower than the average bias of the best performing UBER algorithm, ube-X-leaner. Compared with the traditional double robust estimation, the average deviation is reduced by 38.54%. Referring to table (1) below and fig. 6, fig. 6 is a schematic diagram of a deviation ratio of a processing method of object data provided in an embodiment of the present application. The average bias of the Binary dual robust estimate in fig. 6 is indicated by the dark bars, which have a lower bias than the prior art scheme.
Watch (1)
According to the embodiment of the application, the prediction results respectively corresponding to the intervention object group and the comparison object group are subjected to activation processing and inverse activation processing, so that the prediction results are converted from binary results into continuous results, the value range of the average processing effect is compressed through the activation processing, and the deviation of the existing dual stable estimation is reduced.
Continuing with the exemplary structure of the object data processing device 455 provided by the embodiment of the present application implemented as software modules, in some embodiments, as shown in fig. 2, the software modules stored in the object data processing device 455 of the memory 450 may include: a result prediction module 4551, configured to perform event prediction processing on the intervention object data of the intervention object group and the comparison object data of the comparison object group respectively to obtain a first prediction result of the intervention object group and a second prediction result of the comparison object group, where the event prediction processing is used to predict a response result to an event; an activation processing module 4552 configured to perform activation processing based on the first prediction result and the intervention object data of the intervention object group to obtain a first activation processing result of the intervention object group, and perform activation processing based on the second prediction result and the comparison object data of the comparison object group to obtain a second activation processing result of the comparison object group; a reverse activation processing module 4553, configured to perform reverse activation processing on the first activation processing result and the second activation processing result to obtain a first reverse activation processing result of the intervention object group and a second reverse activation processing result of the comparison object group; a comparison module 4554 configured to use a difference between the first reverse activation processing result of the intervention object group and the second reverse activation processing result of the control object group as an average processing effect between the intervention object group and the control object group; a matching module 4555 configured to determine a similar object of the intervention object group based on the average processing effect when the average processing effect is greater than the average processing effect threshold, and use the similar object as a target object to which the intervention processing is applied.
In some embodiments, the result predicting module 4551 is configured to perform feature extraction on the intervention object data of the intervention object group to obtain intervention object features of the intervention object group, and perform feature extraction on the comparison object data of the comparison object group to obtain comparison object features of the comparison object group; and calling the result prediction model based on the characteristics of the intervention object to obtain a first prediction result of the intervention object group, and calling the result prediction model based on the characteristics of the comparison object to obtain a second prediction result of the comparison object group.
In some embodiments, the result predicting module 4551 is configured to perform conversion processing on the intervention object data of the intervention object group to obtain an object feature vector of the intervention object group, and perform dimension reduction processing on the object feature vector of the intervention object group to obtain an intervention object feature of the intervention object group; and performing conversion processing on the comparison object data of the comparison object group to obtain an object feature vector of the comparison object group, and performing dimension reduction processing on the object feature vector of the comparison object group to obtain the comparison object feature of the comparison object group.
In some embodiments, the intervention object data comprises: the actual response result of the intervention object, the attribute parameters of the intervention object and the intervention data; the control object data includes: comparing the actual response result of the object with the attribute parameters of the object; the outcome prediction model includes: a first outcome prediction model and a second outcome prediction model; the result prediction module 4551 is configured to train the preset logic classification model based on the actual response result of the intervention object, the attribute parameter of the intervention object, and the intervention data to obtain a first result prediction model, and train the preset logic classification model based on the actual response result of the comparison object and the attribute parameter of the comparison object to obtain a second result prediction model.
In some embodiments, the intervention object data comprises: the actual response result of the intervention object, the attribute parameters of the intervention object and the intervention data; the control object data includes: comparing the actual response result of the object with the attribute parameters of the object; an activation processing module 4552 configured to determine a total activation processing result of the intervention object group based on the first prediction result, the intervention probability of each intervention object in the intervention object group, the actual response result of the intervention object, and the intervention data; dividing the total activation processing result of the intervention object group by the number of the intervention objects of the intervention object group to obtain a first activation processing result of the intervention object group; determining a total activation processing result of the control object group based on the second prediction result, the intervention probability of each control object in the control object group and the actual response result of the control object; and performing division processing on the total activation processing result of the control object group and the number of the control objects of the control object group to obtain a second activation processing result of the control object group.
In some embodiments, the activation processing module 4552 is configured to invoke the machine learning classification model to perform the intervention probability prediction processing based on the intervention object attribute parameters of the intervention object group, so as to obtain the intervention probability of each intervention object in the intervention object group, and invoke the machine learning classification model to perform the intervention probability prediction processing based on the control object attribute parameters of the control object group, so as to obtain the intervention probability of each control object in the control object group.
In some embodiments, the activation processing module 4552 is configured to perform the following for each intervention object in the set of intervention objects: determining a first intermediate parameter based on the reciprocal of the first prediction result, and determining a second intermediate parameter based on the reciprocal of the actual response result of the intervention object; taking the ratio of the intervention data to the intervention probability of the intervention object as a third intermediate parameter; acquiring a fourth difference value of the second intermediate parameter and the first intermediate parameter, multiplying the fourth difference value and the third intermediate parameter, adding the product obtained by multiplication and the first intermediate parameter, and taking the sum obtained by addition as an object activation processing result of an intervention object; and adding the object activation processing results of each intervention object to obtain the total activation processing result of the intervention object group.
In some embodiments, the activation processing module 4552 is configured to perform the following for each control subject in the control subject group: determining a fourth intermediate parameter based on the inverse of the second predicted result and a fifth intermediate parameter based on the inverse of the actual response result; acquiring a first difference value between the intervention data of the intervention object corresponding to the probability 1 comparison object, taking the first difference value as the comparison object intervention data of the comparison object, acquiring a second difference value between the intervention probabilities of the probability 1 comparison object, and taking the ratio of the first difference value to the second difference value as a sixth intermediate parameter; acquiring a third difference value between the fifth intermediate parameter and the fourth intermediate parameter, multiplying the third difference value and the sixth intermediate parameter, adding the product obtained by multiplication and the fourth intermediate parameter, and taking the sum obtained by addition as an object activation processing result of a comparison object; and adding the object activation processing results of each comparison object to obtain the total activation processing result of the comparison object group.
In some embodiments, the first activation processing result and the second activation processing result are binary results; the inverse activation processing module 4553 is configured to perform activation processing on the negative number of the first activation processing result as the argument of the activation function to obtain a first inverse activation processing result with a continuous value range, and perform activation processing on the negative number of the second activation processing result as the argument of the activation function to obtain a second inverse activation processing result with a continuous value range.
In some embodiments, the matching module 4555 is configured to obtain an intervention probability of each intervention object in the intervention object group and an intervention probability corresponding to each sample object in the sample object group, wherein each sample object in the sample object group is an object that is not subjected to intervention processing; determining a trend score for each intervention subject based on the intervention probability for each intervention subject, and determining a trend score for each sample subject based on the intervention probability for each sample subject; the following processing is carried out for each intervention object: matching processing is carried out in the sample object group based on the trend score of the intervention object, a sample object with the same trend score as the intervention object is obtained, and the sample object with the same trend score is used as a similar object of the intervention object; and taking each similar object in the sample object group as a target object subjected to the intervention treatment.
In some embodiments, the matching module 4555 is configured to multiply the intervention probability of each intervention object by the intervention parameter to obtain a trend value of each intervention object, round the trend value of each intervention object to obtain a trend score of each intervention object, multiply the intervention probability of each sample object by the intervention parameter to obtain a trend value of each sample object, and round the trend value of each sample object to obtain a trend score of each sample object.
Embodiments of the present application provide a computer program product or computer program comprising computer instructions stored in a computer readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the object data processing method described in the embodiment of the present application.
Embodiments of the present application provide a computer-readable storage medium storing executable instructions, which when executed by a processor, will cause the processor to execute a method for processing object data provided by embodiments of the present application, for example, the method for processing object data as shown in fig. 3A.
In some embodiments, the computer-readable storage medium may be memory such as FRAM, ROM, PROM, EP ROM, EEPROM, flash memory, magnetic surface memory, optical disk, or CD-ROM; or may be various devices including one or any combination of the above memories.
In some embodiments, executable instructions may be written in any form of programming language (including compiled or interpreted languages), in the form of programs, software modules, scripts or code, and may be deployed in any form, including as a stand-alone program or as a module, component, subroutine, or other unit suitable for use in a computing environment.
By way of example, executable instructions may correspond, but do not necessarily have to correspond, to files in a file system, and may be stored in a portion of a file that holds other programs or data, such as in one or more scripts in a hypertext Markup Language (H TML) document, in a single file dedicated to the program in question, or in multiple coordinated files (e.g., files that store one or more modules, sub-programs, or portions of code).
By way of example, executable instructions may be deployed to be executed on one computing device or on multiple computing devices at one site or distributed across multiple sites and interconnected by a communication network.
In summary, according to the embodiment of the present application, the prediction results corresponding to the intervention object group and the comparison object group are respectively subjected to activation processing and inverse activation processing, so that the prediction results are converted from binary results into continuous results, the value range of the average processing effect is compressed through the activation processing, the deviation of the double steady estimation is reduced, the accuracy of determining the target object is improved, and the effectiveness of applying the intervention processing is improved.
The above description is only an example of the present application, and is not intended to limit the scope of the present application. Any modification, equivalent replacement, and improvement made within the spirit and scope of the present application are included in the protection scope of the present application.
Claims (15)
1. A method for processing object data, the method comprising:
respectively carrying out event prediction processing on intervention object data of an intervention object group and comparison object data of a comparison object group to obtain a first prediction result of the intervention object group and a second prediction result of the comparison object group, wherein the event prediction processing is used for predicting a response result aiming at an event;
performing activation processing on the basis of the first prediction result and the intervention object data of the intervention object group to obtain a first activation processing result of the intervention object group, and performing activation processing on the basis of the second prediction result and the control object data of the control object group to obtain a second activation processing result of the control object group;
performing reverse activation processing on the first activation processing result and the second activation processing result to obtain a first reverse activation processing result of the intervention object group and a second reverse activation processing result of the control object group;
taking the difference value between the first reverse activation processing result of the intervention object group and the second reverse activation processing result of the control object group as the average processing effect between the intervention object group and the control object group;
when the average treatment effect is larger than an average treatment effect threshold value, determining similar objects of the intervention object group based on the average treatment effect, and regarding the similar objects as target objects subjected to intervention treatment.
2. The method of claim 1, wherein the performing event prediction processing on the intervention object data of the intervention object group and the control object data of the control object group to obtain a first prediction result of the intervention object group and a second prediction result of the control object group respectively comprises:
performing feature extraction on the intervention object data of the intervention object group to obtain intervention object features of the intervention object group, and performing feature extraction on the comparison object data of the comparison object group to obtain comparison object features of the comparison object group;
and calling the result prediction model based on the characteristics of the control object to obtain a second prediction result of the control object group.
3. The method of claim 2, wherein the step of performing feature extraction on the intervention object data of the intervention object group to obtain the intervention object features of the intervention object group and performing feature extraction on the control object data of the control object group to obtain the control object features of the control object group comprises:
performing conversion processing on the intervention object data of the intervention object group to obtain an object feature vector of the intervention object group, and performing dimension reduction processing on the object feature vector of the intervention object group to obtain an intervention object feature of the intervention object group;
and performing conversion processing on the comparison object data of the comparison object group to obtain an object feature vector of the comparison object group, and performing dimension reduction processing on the object feature vector of the comparison object group to obtain the comparison object feature of the comparison object group.
4. The method of claim 2,
the intervention object data comprises: the actual response result of the intervention object, the attribute parameters of the intervention object and the intervention data; the control object data includes: comparing the actual response result of the object with the attribute parameters of the object;
the outcome prediction model includes: a first outcome prediction model and a second outcome prediction model;
before the obtaining of the first prediction result of the intervention object group based on the intervention object feature calling result prediction model and the obtaining of the second prediction result of the control object group based on the control object feature calling result prediction model, the method further includes:
training a preset logic classification model based on the actual response result of the intervention object, the attribute parameters of the intervention object and the intervention data to obtain the first result prediction model, an
And training the preset logic classification model based on the comparison object actual response result and the comparison object attribute parameters to obtain the second result prediction model.
5. The method of claim 1,
the intervention object data comprises: the actual response result of the intervention object, the attribute parameters of the intervention object and the intervention data; the control object data includes: comparing the actual response result of the object with the attribute parameters of the object;
the step of performing activation processing based on the first prediction result and the intervention object data of the intervention object group to obtain a first activation processing result of the intervention object group, and performing activation processing based on the second prediction result and the control object data of the control object group to obtain a second activation processing result of the control object group includes:
acquiring the intervention probabilities respectively corresponding to the intervention object group and the control object group;
determining a total activation processing outcome for the set of intervention subjects based on the first prediction outcome, the intervention probability for each intervention subject in the set of intervention subjects, the actual response outcome for the intervention subject, and the intervention data;
dividing the total activation processing result of the intervention object group by the number of the intervention objects of the intervention object group to obtain a first activation processing result of the intervention object group;
determining a total activation treatment outcome for the group of control subjects based on the second predicted outcome, the intervention probability for each control subject in the group of control subjects, and the actual response outcome for the control subject;
and performing division processing on the total activation processing result of the control object group and the number of the control objects of the control object group to obtain a second activation processing result of the control object group.
6. The method of claim 5, wherein said obtaining the intervention probabilities corresponding to said group of intervention subjects and said group of control subjects, respectively, comprises:
based on the intervention object attribute parameters of the intervention object group, calling a machine learning classification model to perform intervention probability prediction processing to obtain the intervention probability of each intervention object in the intervention object group, and
and calling the machine learning classification model to perform intervention probability prediction processing based on the attribute parameters of the control objects in the control object group to obtain the intervention probability of each control object in the control object group.
7. The method of claim 5, wherein said determining a total activation treatment outcome for said set of intervention subjects based on said first predicted outcome, an intervention probability for each intervention subject in said set of intervention subjects, an actual response outcome for said intervention subject, and said intervention data comprises:
for each intervention object in the group of intervention objects, performing the following:
determining a first intermediate parameter based on the inverse of the first prediction result, and determining a second intermediate parameter based on the inverse of the actual response result of the intervention object;
taking the ratio of the intervention data to the intervention probability of the intervention object as a third intermediate parameter;
acquiring a fourth difference value between the second intermediate parameter and the first intermediate parameter, multiplying the fourth difference value by the third intermediate parameter, adding the product obtained by the multiplication to the first intermediate parameter, and taking the sum obtained by the addition as an object activation processing result of the intervention object;
and adding the object activation processing results of each intervention object to obtain the total activation processing result of the intervention object group.
8. The method of claim 5, wherein said determining a total activation treatment outcome for said group of control subjects based on said second prediction outcome, the probability of intervention for each control subject in said group of control subjects, and the actual response outcome for said control subjects comprises:
performing the following for each of the control subjects in the group of control subjects:
determining a fourth intermediate parameter based on the inverse of the second predicted outcome and a fifth intermediate parameter based on the inverse of the actual response outcome;
obtaining a first difference value between the intervention data of the intervention object corresponding to the comparison object with the probability 1, taking the first difference value as the intervention data of the comparison object, obtaining a second difference value between the intervention probabilities of the comparison object with the probability 1, and taking the ratio of the first difference value to the second difference value as a sixth intermediate parameter;
acquiring a third difference value between the fifth intermediate parameter and the fourth intermediate parameter, multiplying the third difference value by the sixth intermediate parameter, adding the product obtained by the multiplication to the fourth intermediate parameter, and taking the sum obtained by the addition as an object activation processing result of the comparison object;
and adding the object activation processing results of each control object to obtain the total activation processing result of the control object group.
9. The method of claim 1,
the first activation processing result and the second activation processing result are binary results;
the performing reverse activation processing on the first activation processing result and the second activation processing result to obtain a first reverse activation processing result of the intervention object group and a second reverse activation processing result of the control object group, includes:
activating the negative number of the first activation processing result as an independent variable of an activation function to obtain a first reverse activation processing result with a continuous value range, and
and taking the negative number of the second activation processing result as the independent variable of the activation function to carry out activation processing, and obtaining a second reverse activation processing result with a continuous value range.
10. The method of claim 1, wherein said determining similar objects of said group of intervening objects based on said average treatment effect and having said similar objects as target objects for which intervening treatment is performed comprises:
acquiring the intervention probability of each intervention object in the intervention object group and the intervention probability corresponding to each sample object in a sample object group, wherein each sample object in the sample object group is an object which is not subjected to intervention treatment;
determining a trend score for each of the intervention subjects based on the intervention probability for each of the intervention subjects and a trend score for each of the sample subjects based on the intervention probability for each of the sample subjects;
for each of the intervention objects, performing the following:
matching processing is carried out in the sample object group based on the trend score of the intervention object, a sample object with the same trend score as the intervention object is obtained, and the sample object with the same trend score is used as a similar object of the intervention object;
and taking each similar object in the sample object group as a target object subjected to intervention treatment.
11. The method of claim 10 wherein said determining a trend score for each of said intervention subjects based on their intervention probability and determining a trend score for each of said sample subjects based on their intervention probability comprises:
multiplying the intervention probability of each intervention object by the intervention parameters to obtain a trend value of each intervention object, rounding the trend value of each intervention object to obtain a trend score of each intervention object, and
and multiplying the intervention probability of each sample object by the intervention parameter to obtain a trend value of each sample object, and rounding the trend value of each sample object to obtain a trend score of each sample object.
12. An apparatus for processing object data, the apparatus comprising:
the result prediction module is configured to perform event prediction processing on intervention object data of an intervention object group and control object data of a control object group respectively to obtain a first prediction result of the intervention object group and a second prediction result of the control object group, wherein the event prediction processing is used for predicting a response result to an event;
an activation processing module configured to perform activation processing based on the first prediction result and the intervention object data of the intervention object group to obtain a first activation processing result of the intervention object group, and perform activation processing based on the second prediction result and the control object data of the control object group to obtain a second activation processing result of the control object group;
a reverse activation processing module configured to perform reverse activation processing on the first activation processing result and the second activation processing result to obtain a first reverse activation processing result of the intervention object group and a second reverse activation processing result of the control object group;
a comparison module configured to use a difference between a first reverse activation processing result of the intervention object group and a second reverse activation processing result of the control object group as an average processing effect between the intervention object group and the control object group;
and the matching module is configured to determine similar objects of the intervention object group based on the average treatment effect when the average treatment effect is greater than an average treatment effect threshold value, and take the similar objects as target objects subjected to intervention treatment.
13. An electronic device, characterized in that the electronic device comprises:
a memory for storing executable instructions;
a processor for implementing the method of any one of claims 1 to 11 when executing executable instructions stored in the memory.
14. A computer-readable storage medium storing executable instructions, wherein the executable instructions, when executed by a processor, implement the method of any one of claims 1 to 11.
15. A computer program product comprising a computer program or instructions, characterized in that the computer program or instructions, when executed by a processor, implement the method of any of claims 1 to 11.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210555254.4A CN115115093A (en) | 2022-05-19 | 2022-05-19 | Object data processing method and device, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210555254.4A CN115115093A (en) | 2022-05-19 | 2022-05-19 | Object data processing method and device, electronic equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115115093A true CN115115093A (en) | 2022-09-27 |
Family
ID=83326207
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210555254.4A Pending CN115115093A (en) | 2022-05-19 | 2022-05-19 | Object data processing method and device, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115115093A (en) |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190188594A1 (en) * | 2017-12-18 | 2019-06-20 | Microsoft Technology Licensing, Llc | Predicting site visit based on intervention |
| CN111027714A (en) * | 2019-12-11 | 2020-04-17 | 腾讯科技（深圳）有限公司 | Artificial intelligence-based object recommendation model training method, recommendation method and device |
| US20200125990A1 (en) * | 2018-10-23 | 2020-04-23 | Google Llc | Systems and Methods for Intervention Optimization |
| CN112085281A (en) * | 2020-09-11 | 2020-12-15 | 支付宝(杭州)信息技术有限公司 | Method and device for detecting safety of business prediction model |
| CN112257470A (en) * | 2020-11-12 | 2021-01-22 | 腾讯科技（深圳）有限公司 | Model training method and device, computer equipment and readable storage medium |
| CN112825576A (en) * | 2019-11-20 | 2021-05-21 | 中国电信股份有限公司 | Method and device for determining cell capacity expansion and storage medium |
| CN112883264A (en) * | 2021-02-09 | 2021-06-01 | 联想(北京)有限公司 | Recommendation method and device |
| CN113704637A (en) * | 2021-08-30 | 2021-11-26 | 深圳前海微众银行股份有限公司 | Object recommendation method, device and storage medium based on artificial intelligence |
| JP2021184173A (en) * | 2020-05-22 | 2021-12-02 | Ｓｍｎ株式会社 | Apparatus, method and program for calculating reward corresponding to intervention effect |
| CN114139724A (en) * | 2021-11-30 | 2022-03-04 | 支付宝(杭州)信息技术有限公司 | Method and device for training gain model |
-
2022
- 2022-05-19 CN CN202210555254.4A patent/CN115115093A/en active Pending
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190188594A1 (en) * | 2017-12-18 | 2019-06-20 | Microsoft Technology Licensing, Llc | Predicting site visit based on intervention |
| US20200125990A1 (en) * | 2018-10-23 | 2020-04-23 | Google Llc | Systems and Methods for Intervention Optimization |
| CN112825576A (en) * | 2019-11-20 | 2021-05-21 | 中国电信股份有限公司 | Method and device for determining cell capacity expansion and storage medium |
| CN111027714A (en) * | 2019-12-11 | 2020-04-17 | 腾讯科技（深圳）有限公司 | Artificial intelligence-based object recommendation model training method, recommendation method and device |
| JP2021184173A (en) * | 2020-05-22 | 2021-12-02 | Ｓｍｎ株式会社 | Apparatus, method and program for calculating reward corresponding to intervention effect |
| CN112085281A (en) * | 2020-09-11 | 2020-12-15 | 支付宝(杭州)信息技术有限公司 | Method and device for detecting safety of business prediction model |
| CN112257470A (en) * | 2020-11-12 | 2021-01-22 | 腾讯科技（深圳）有限公司 | Model training method and device, computer equipment and readable storage medium |
| CN112883264A (en) * | 2021-02-09 | 2021-06-01 | 联想(北京)有限公司 | Recommendation method and device |
| CN113704637A (en) * | 2021-08-30 | 2021-11-26 | 深圳前海微众银行股份有限公司 | Object recommendation method, device and storage medium based on artificial intelligence |
| CN114139724A (en) * | 2021-11-30 | 2022-03-04 | 支付宝(杭州)信息技术有限公司 | Method and device for training gain model |
Non-Patent Citations (1)
| Title |
|---|
| NEELAM YOUNAS ET.AL.: "Optimal Causal Decision Trees Ensemble for Improved Prediction and Causal Inference", 《IEEE ACCESS》, vol. 10, 19 January 2022 (2022-01-19), pages 13000 - 13011 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Li et al. | Project success prediction in crowdfunding environments | |
| CN104574192B (en) | Method and device for identifying same user in multiple social networks | |
| Linton et al. | Dynamic topic modelling for cryptocurrency community forums | |
| US20190392258A1 (en) | Method and apparatus for generating information | |
| de Sá et al. | Label ranking forests | |
| US20170235726A1 (en) | Information identification and extraction | |
| US11874798B2 (en) | Smart dataset collection system | |
| CN111369344B (en) | Method and device for dynamically generating early warning rules | |
| Woo | Market basket analysis algorithms with mapreduce | |
| Wu et al. | Collaborative filtering recommendation based on conditional probability and weight adjusting | |
| CN109636627B (en) | Insurance product management method, device, medium and electronic equipment based on block chain | |
| CN111667018A (en) | Object clustering method and device, computer readable medium and electronic equipment | |
| Huang et al. | An adapted firefly algorithm for product development project scheduling with fuzzy activity duration | |
| CN115115093A (en) | Object data processing method and device, electronic equipment and storage medium | |
| Wang | Is Human Culture Locked by Evolution? | |
| CN113742495B (en) | Rating feature weight determining method and device based on prediction model and electronic equipment | |
| Trancik | Testing and improving technology forecasts for better climate policy | |
| Chang et al. | A novel approach for rumor detection in social platforms: Memory-augmented transformer with graph convolutional networks | |
| CA3097731A1 (en) | System and method for deep learning recommender | |
| Qing et al. | Data of e-commerce users based on data mining technology | |
| CN117574410B (en) | Risk data detection method and device | |
| CN115222486B (en) | Article recommendation model training method, article recommendation method, device and storage medium | |
| Zhang et al. | The spread of information in virtual communities | |
| Nescolarde-Selva et al. | Synonymy relationship and stochastic processes in determination of flow equations in ecological models | |
| Raikwal et al. | Weight based classification algorithm for medical data |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |