WO2017193897A1 - 一种数据推荐方法及其设备、存储介质 - Google Patents

一种数据推荐方法及其设备、存储介质 Download PDF

Info

Publication number
WO2017193897A1
WO2017193897A1 PCT/CN2017/083524 CN2017083524W WO2017193897A1 WO 2017193897 A1 WO2017193897 A1 WO 2017193897A1 CN 2017083524 W CN2017083524 W CN 2017083524W WO 2017193897 A1 WO2017193897 A1 WO 2017193897A1
Authority
WO
WIPO (PCT)
Prior art keywords
service
attribute
data
service data
terminal identifier
Prior art date
Application number
PCT/CN2017/083524
Other languages
English (en)
French (fr)
Inventor
黄安埠
Original Assignee
腾讯科技(深圳)有限公司
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 腾讯科技(深圳)有限公司 filed Critical 腾讯科技(深圳)有限公司
Publication of WO2017193897A1 publication Critical patent/WO2017193897A1/zh
Priority to US16/049,352 priority Critical patent/US10706363B2/en

Links

Images

Classifications

    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L67/00Network arrangements or protocols for supporting network services or applications
    • H04L67/50Network services
    • H04L67/55Push-based network services
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00Computing arrangements using knowledge-based models
    • G06N5/04Inference or reasoning models
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/04Architecture, e.g. interconnection topology
    • G06N3/047Probabilistic or stochastic networks
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00Computing arrangements using knowledge-based models
    • G06N5/04Inference or reasoning models
    • G06N5/046Forward inferencing; Production systems
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION 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
    • G06Q30/00Commerce
    • G06Q30/02Marketing; Price estimation or determination; Fundraising
    • G06Q30/0282Rating or review of business operators or products
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION 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
    • G06Q30/00Commerce
    • G06Q30/06Buying, selling or leasing transactions
    • G06Q30/0601Electronic shopping [e-shopping]
    • G06Q30/0631Item recommendations
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L67/00Network arrangements or protocols for supporting network services or applications
    • H04L67/2866Architectures; Arrangements
    • H04L67/30Profiles
    • H04L67/306User profiles
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00Machine learning
    • G06N20/20Ensemble learning
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/04Architecture, e.g. interconnection topology
    • G06N3/044Recurrent networks, e.g. Hopfield networks
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00Computing arrangements using knowledge-based models
    • G06N5/01Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound

Definitions

  • the present invention relates to the field of computer technologies, and in particular, to a data recommendation method, a device thereof, and a storage medium.
  • the background server not only ensures the normal operation of the terminal application, but also recommends related business data, for example, recommending daily featured songs in the music application. Recommended hot items in the shopping app, which enhances the user experience.
  • the existing recommendation architecture is composed of an offline layer and a real-time layer.
  • the offline layer is responsible for counting the behaviors of the user for the current service during a period of time, thereby calculating the tag attributes, similar data, and other timed service attributes of each user.
  • the real-time layer is responsible for re-screening the recommended business data, and outputs the filtered business data to the user terminal, and also needs to perform feedback learning according to the user's real-time behavioral operation for the current business, Generate real-time business attributes such as real-time tag attributes for each user.
  • the real-time layer real-time feedback learning service attribute often needs to wait for the offline layer at the next time.
  • the recommended business data is selected, it can take effect. Therefore, the recommended business data cannot be adjusted in real time, which affects the effect of the business data recommendation.
  • the embodiment of the present invention provides a data recommendation method, a device, and a storage medium, which can generate recommended service data in real time based on service attributes provided by an offline layer and a real-time layer, improve update efficiency of recommended service data, and improve service data recommendation effect. .
  • the embodiment of the invention provides a data recommendation method, which may include:
  • timing service attribute corresponding to the terminal identifier of the user terminal and a real-time service attribute corresponding to the terminal identifier, where the timing service attribute is a first service for the target service uploaded according to the terminal identifier in the first preset time period
  • first label attribute corresponding to the terminal identifier is calculated
  • real-time service attribute is a second service operation for the target service that is uploaded in real time according to the terminal identifier
  • the calculated terminal identifier corresponds to Second tag attribute
  • the embodiment of the invention provides a data recommendation device, which may include:
  • the attribute obtaining unit is configured to acquire a timed service attribute corresponding to the terminal identifier of the user terminal and a real-time service attribute corresponding to the terminal identifier, where the timed service attribute is uploaded according to the terminal identifier in the first preset time period. a first service operation of the target service, where the first tag attribute corresponding to the terminal identifier is calculated; the real-time service attribute is a second service operation for the target service uploaded in real time according to the terminal identifier, and the calculated The second tag attribute corresponding to the terminal identifier;
  • a recommendation data selection unit configured to select at least one recommended service data in the service data set corresponding to the target service according to the timing service attribute and the real-time service attribute;
  • a data sending unit configured to send the at least one recommended service data to the user terminal.
  • the embodiment of the present invention provides a computer storage medium, where the computer storage medium stores computer executable instructions, and the computer executable instructions are used to execute the data recommendation method provided by the embodiment of the present invention.
  • An embodiment of the present invention provides a data recommendation device, including:
  • a storage medium configured to store executable instructions
  • a processor configured to execute the stored executable instructions for performing the following steps:
  • timing service attribute corresponding to the terminal identifier of the user terminal and a real-time service attribute corresponding to the terminal identifier, where the timing service attribute is a first service for the target service uploaded according to the terminal identifier in the first preset time period
  • first label attribute corresponding to the terminal identifier is calculated
  • real-time service attribute is a second service operation for the target service that is uploaded in real time according to the terminal identifier
  • the calculated terminal identifier corresponds to Second tag attribute
  • the recommended service data is selected in the service data set in real time by acquiring the timing service attribute of the offline layer and the real-time service attribute of the real-time layer, and is pushed to the user terminal.
  • the process of realizing the recommended service data based on the service attributes provided by the offline layer and the real-time layer is realized, the update efficiency of the recommended service data is improved, the effect of the service data recommendation is improved, and the business data selection and real-time of the offline layer are shared.
  • the work of the layer's business data recommendation improves the efficiency of business data recommendation.
  • FIG. 1 is a schematic flowchart of a data recommendation method according to an embodiment of the present invention.
  • FIG. 2 is a schematic flowchart of another data recommendation method according to an embodiment of the present invention.
  • FIG. 3 is a schematic diagram of an example of a system for recommending data according to an embodiment of the present invention.
  • FIG. 4 is a schematic structural diagram of a data recommendation device according to an embodiment of the present invention.
  • FIG. 5 is a schematic structural diagram of a recommendation data selection unit according to an embodiment of the present invention.
  • FIG. 6 is a schematic structural diagram of a data sending unit according to an embodiment of the present invention.
  • FIG. 7 is a schematic structural diagram of another data recommendation device according to an embodiment of the present invention.
  • the data recommendation method provided by the embodiment of the present invention may be applied to a scenario in which music data is recommended.
  • the data recommendation device acquires a timed music tag attribute corresponding to the terminal identifier of the user terminal in the music application, and a real-time music tag attribute corresponding to the terminal identifier.
  • the data recommendation device selects at least one recommended music data in the music data set according to the timed music tag attribute and the real-time music tag attribute, and the data recommendation device sends the at least one recommended music data to the
  • the scene of the user terminal is also applicable to the scenario in which the product data is recommended.
  • the data recommendation device acquires the terminal identifier of the user terminal, the corresponding timed product label attribute in the online shopping application, and the real-time product label attribute corresponding to the terminal identifier.
  • the data recommendation device purchases the online product according to the timed item label attribute and the real-time item label attribute
  • At least one recommended product data is selected from the corresponding product data set, and the data recommendation device sends the at least one recommended product data to a scene of the user terminal or the like.
  • the process of realizing the recommended service data based on the service attributes provided by the offline layer and the real-time layer is realized, the update efficiency of the recommended service data is improved, the effect of the service data recommendation is improved, and the business data selection and real-time of the offline layer are shared.
  • the work of the layer's business data recommendation improves the efficiency of business data recommendation.
  • the data recommendation device may be a background service device corresponding to the target service, configured to generate corresponding recommended service data according to the tag attributes of different dimensions corresponding to the terminal identifier of the user terminal, and push the corresponding recommended service data to the user terminal;
  • the terminal may include a terminal device such as a tablet computer, a smart phone, a palmtop computer, and a mobile internet device (MID) having functions such as supporting a target service;
  • the target service may be an application that provides a business requirement, such as a music application or a movie application. , online shopping applications, etc.
  • FIG. 1 is a schematic flowchart diagram of a data recommendation method according to an embodiment of the present invention. As shown in FIG. 1, the method of the embodiment of the present invention may include the following steps S101 to S103.
  • the data recommendation device may obtain the terminal identifier of the user terminal for the corresponding scheduled service attribute in the target service and the real-time service attribute corresponding to the terminal identifier, and it should be noted that the user terminal
  • the terminal identifier may be a unique identifier of the user terminal, or may be a user registration of the target service in advance, and use the user terminal to log in to the target service, for example, an application account of the music application, Application account for online shopping applications, etc.
  • the timing service attribute is based on the terminal identifier in the first pre-
  • the first service attribute corresponding to the terminal identifier is calculated by using the first service operation for the target service that is uploaded in the time period.
  • the timed service attribute may be the terminal identifier based on the record.
  • the first label attribute is obtained by calculation of the offline layer, for example, for the music application, acquiring the first service operation for the song in the music application uploaded by the terminal identifier within one month
  • the operation includes downloading, skipping, deleting, collecting, etc.
  • the offline layer may adopt a preset recommendation algorithm model, for example, a user-based (Base-Based) and an Item-Based collaborative filtering algorithm model, based on content.
  • the first tag attribute corresponding to the terminal identifier includes a comprehensive music genre, a singer preference, an age preference, and the like within one month, and may also calculate other music data and the like similar to the music list under the terminal identifier;
  • the real-time service attribute is a second label attribute corresponding to the terminal identifier calculated by the second service operation for the target service that is uploaded in real time according to the terminal identifier.
  • the real-time service attribute may be And the second label attribute corresponding to the terminal identifier calculated by using the preset confidence model, the second label attribute corresponding to the terminal identifier calculated by the terminal identifier in the second preset time period, A preset time period is greater than the second preset time period.
  • the second tag attribute is obtained by calculating the real-time layer, for example, for a relatively static application scenario such as a daily selection in a music application, acquiring the target that the terminal identifier is uploaded within 1 day.
  • the second business operation of the music data in the music application including downloading, skipping, deleting, collecting, and the like.
  • the real-time layer may adopt a preset confidence model, for example, a Stochastic Gradient Descent (SGD) algorithm model and a Follow The Regularized Leader (FTRL) algorithm model.
  • SGD Stochastic Gradient Descent
  • FTRL Repeat The Regularized Leader
  • calculating a second tag attribute corresponding to the terminal identifier including a comprehensive music genre, a singer preference, an age preference, and the like within one day.
  • the real-time service attribute may be a second service operation based on the recorded terminal identifier to the historical recommended service data, and the terminal identifier corresponding to the preset confidence model is used to calculate
  • the real-time service attribute, the historical recommendation service data is the last batch of recommended service data based on the recommended service data, for example, for a dynamic application scenario such as guessing your favorite in the music application, 15 music data may be selected at one time as A batch of recommended service data is sent in three times, each time for 5 times. Therefore, the second service operation of the recommended service data (ie, historical recommended service data) uploaded by the terminal identifier may be obtained, including downloading and skipping. , delete, collect, etc.
  • the real-time layer may also adopt a preset confidence model, and calculate a second label attribute corresponding to the terminal identifier according to the second service operation in the terminal identifier, including the second service operation of the recommended service data of the batch.
  • first service operation and the second service operation may be the same service operation, and the first and second divisions are only used to describe the difference between the timing and the real-time.
  • the data recommendation device may select at least one recommended service data in the service data set corresponding to the target service according to the timing service attribute and the real-time service attribute.
  • the service data set includes at least one service data, and each of the at least one service data has a third tag attribute corresponding to each of the service data, for example, for music.
  • the application, the music data in the music library corresponds to a third tag attribute such as a music genre, a singer, and a genre, and the data recommendation device may obtain the third tag attribute in the at least one service data to satisfy the timing service.
  • the data recommendation device may Selecting, in the service data set, all the service data that meets the timing service attribute and the real-time service attribute by the third tag attribute, the at least one target service data representation according to the timed service attribute and the real-time All business data selected by the business attribute in the business data set.
  • the data recommendation device adopts a preset scoring model according to the timing service attribute and the real-time service attribute, and the preset scoring model may be a Gradient Boosting Decision Tree (GBDT) algorithm model, and acquire the at least one target.
  • GBDT Gradient Boosting Decision Tree
  • the data recommendation device may send the at least one recommended service data to the user terminal corresponding to the terminal identifier. In other embodiments of the present invention, the data recommendation device may search for the terminal identifier when receiving the service data recommendation request that is sent by the user terminal and carrying the terminal identifier.
  • the data recommendation device may be configured according to each of the at least one recommended service data And the at least one recommended service data is subjected to a sorting process, and after the sorting process, the at least one recommended service data is divided into at least one recommended service data set, the at least one Each recommended service data set in the recommended service data set includes a preset number of recommended service data. For example, for the music application, a total of 15 pieces of music data are selected as a batch of recommended service data, and are divided into 3 times, and 5 times are pushed each time.
  • the data recommendation device may sequentially send the at least one recommended service data set to the user terminal.
  • a recommendation may be sent every time a service data recommendation request is received. After the service data recommendation request is received, the recommended service data set is sent at intervals. It can be understood that the recommended service data set sent this time is different from the recommended service data set sent by the history.
  • the step of receiving the service data recommendation request may be performed before the step of acquiring the timing service attribute corresponding to the terminal identifier of the user terminal and the real-time service attribute corresponding to the terminal identifier, that is, receiving the service data recommendation request, and then performing the recommended service.
  • the process of selecting data may be determined according to the service requirements of different target services.
  • the recommended service data is selected in the service data set in real time by acquiring the timing service attribute of the offline layer and the real-time service attribute of the real-time layer, and is pushed to the user terminal.
  • the process of realizing the recommended service data based on the service attributes provided by the offline layer and the real-time layer is realized, the update efficiency of the recommended service data is improved, the effect of the service data recommendation is improved, and the business data selection and real-time of the offline layer are shared.
  • the work of the layer's business data recommendation improves the efficiency of business data recommendation.
  • FIG. 2 is a schematic flowchart diagram of another data recommendation method according to an embodiment of the present invention. As shown in FIG. 2, the method of the embodiment of the present invention may include the following steps S201 to S207.
  • the data recommendation device may obtain the terminal identifier of the user terminal for the corresponding scheduled service attribute in the target service and the real-time service attribute corresponding to the terminal identifier, and it should be noted that the user terminal
  • the terminal identifier may be a unique identifier of the user terminal, or may be a user registration of the target service in advance, and use the user terminal to log in to the target service, for example, an application account of the music application, Application account for online shopping applications, etc.
  • the timing service attribute is based on the terminal identifier in the first pre-
  • the first service attribute corresponding to the terminal identifier is calculated by using the first service operation for the target service that is uploaded in the time period. Further, the timed service attribute may be the terminal identifier based on the record.
  • the first label attribute is obtained by calculation of the offline layer, for example, for the music application, acquiring the first service operation for the song in the music application uploaded by the terminal identifier within one month
  • the operation includes downloading, skipping, deleting, and collecting operations.
  • the offline layer may adopt a preset recommendation algorithm model, such as a User-Based and Item-Based collaborative filtering algorithm model, a Content-Based recommendation algorithm model, an RBM model, An RNN model or the like, and calculating, according to the first operation under the terminal identifier, a first tag attribute corresponding to the terminal identifier, including a comprehensive music genre, a singer preference, an age preference, etc. within one month, and calculating the terminal Other similar music data, etc. of the music list under the logo;
  • a preset recommendation algorithm model such as a User-Based and Item-Based collaborative filtering algorithm model, a Content-Based recommendation algorithm model, an RBM model, An RNN model or the like
  • the real-time service attribute is a second label attribute corresponding to the terminal identifier calculated by the second service operation for the target service that is uploaded in real time according to the terminal identifier.
  • the real-time service attribute may be And the second label attribute corresponding to the terminal identifier calculated by using the preset confidence model, the second label attribute corresponding to the terminal identifier calculated by the terminal identifier in the second preset time period, A preset time period is greater than the second preset time period.
  • the second tag attribute is obtained by calculating the real-time layer, for example, for a relatively static application scenario such as a daily selection in a music application, acquiring the target that the terminal identifier is uploaded within 1 day.
  • the second business operation of the music data in the music application including downloading, skipping, deleting, collecting, and the like.
  • the real-time layer may adopt a preset confidence model, for example, an SGD algorithm model, an FTRL algorithm model, etc., and calculate a second label attribute corresponding to the terminal identifier according to the second service operation of the terminal identifier, including one day.
  • a preset confidence model for example, an SGD algorithm model, an FTRL algorithm model, etc.
  • the real-time service attribute may be based on a record
  • the terminal identifies the real-time service attribute corresponding to the terminal identifier that is calculated by using the preset confidence model, and the historical recommendation service data is the previous batch based on the recommended service data.
  • Recommend business data for example, for dynamic application scenarios such as guessing your favorite in music applications, you can select 15 music data as a batch of recommended business data at one time, and divide it 3 times, 5 times each time, so you can get
  • the terminal identifies a second service operation that is uploaded to the recommended service data of the batch (ie, historical recommended service data), including operations such as downloading, skipping, deleting, and collecting.
  • the real-time layer may also adopt a preset confidence model, and calculate a second label attribute corresponding to the terminal identifier according to the second service operation in the terminal identifier, including the second service operation of the recommended service data of the batch.
  • first service operation and the second service operation may be the same service operation, and the first and second divisions are only used to describe the difference between the timing and the real-time.
  • the service data set includes at least one service data, and each of the at least one service data has a third tag attribute corresponding to each of the service data, for example, for music.
  • the application the music data in the music library corresponds to a third tag attribute such as a music genre, a singer, and a genre, and the data recommendation device may obtain the third tag attribute in the at least one service data to satisfy the timing service.
  • Attributes and at least one target service data of the real-time service attribute wherein the data recommendation device may select, in the service data set, all service data that the third tag attribute meets the timing service attribute and the real-time service attribute
  • the at least one target service data represents all service data selected in the service data set according to the timing service attribute and the real-time service attribute.
  • the data recommendation device may acquire, in the at least one service data, a first candidate industry corresponding to a third tag attribute that matches the timing service attribute.
  • Data for example, for music applications, A music data exists in the music data set, and the third tag attribute may be XX genre, YY singer, ZZ era, B music data exists, and the third tag attribute may be OO genre, In the PP singer and QQ era, there are C music data, and the third tag attribute can be DD genre, EE singer, FF era, the timing service attribute is XX genre and QQ age, then the data recommendation device can put A music data And B music data as the first candidate business data.
  • the data recommendation device acquires second candidate service data corresponding to the third tag attribute that matches the real-time service attribute in the at least one service data, and according to the above example, if the real-time layer detects the execution of the terminal identifier
  • the second business operation includes continuously collecting the music data of the EE singer, and the data recommendation device may use the C music data as the second candidate business data.
  • the data recommendation device determines the first candidate service data and the second candidate service data as at least one target service data.
  • the data recommendation device may further acquire, in the at least one service data, third candidate service data corresponding to the third tag attribute that matches the timing service attribute, for example, for a music application.
  • the third tag attribute can be XX genre, YY singer, ZZ era
  • B music data exists
  • the third tag attribute can be OO genre, PP singer, QQ age, presence C
  • the music data, the third tag attribute may be DD genre, EE singer, FF era, the timed service attribute is XX genre, PP singer and FF era
  • the data recommendation device may be A music data, B music data and C music data is used as the third candidate business data.
  • the data recommendation device deletes the fourth candidate service data corresponding to the third tag attribute that matches the real-time service attribute in the third candidate service data, and according to the above example, if the real-time layer detects the second execution of the terminal identifier
  • the business operation includes deleting the music data of the PP singer continuously, and the data recommendation device may use the B music data as the fourth candidate service data, and the data recommendation device is based on the third candidate that deleted the fourth candidate service data.
  • the business data generates at least one target business data.
  • the data recommendation device adopts a preset scoring model according to the timing service attribute and the real-time service attribute, and the preset scoring model may be a GBDT algorithm model, and the at least one target service is acquired. a score value of each target service data in the data. Further, the data recommendation device may use the timed service attribute and the real-time service attribute as weight parameters of the preset scoring model to calculate the at least one target service. The score value of each target business data in the data.
  • the data recommendation device acquires at least one recommended service data in the at least one target service data according to the size of the score value of each target service data. In other embodiments of the present invention, the data recommendation device may sort the at least one target service data according to the size of the score value of each target service data, and sort the at least one target service data. Selecting a preset number of target service data with the highest score value as at least one recommended service data.
  • the data recommendation device may further perform re-screening on the selected preset number of the target service data with the highest score value based on the preset screening rule to obtain the at least one recommended service data
  • the screening rule may be a screening rule set by the developer based on the service characteristics of the target service, for example, for a music application, the screening rule may specify that the selected preset number of music data with the highest score value cannot exist in the same singer's label attribute, etc. .
  • the data recommendation device may send the at least one recommended service data to the user terminal corresponding to the terminal identifier.
  • the data recommendation device may search for the terminal identifier when receiving the service data recommendation request that is sent by the user terminal and carrying the terminal identifier.
  • the at least one recommended service data is identified, and the at least one recommended service data is sent to the user terminal.
  • the at least one recommended service data is divided into at least one recommended service data set according to the at least one recommended service data after the sorting process.
  • the data recommendation device may sort the at least one recommended service data according to a score value of each recommended service data in the at least one recommended service data, and after sorting And dividing the at least one recommended service data into at least one recommended service data set, where each recommended service data set in the at least one recommended service data set includes a preset number of recommended service data, for example, for music Application, a total of 15 pieces of music data are selected as a batch of recommended service data, and are divided into 3 times, each time for 5 pushes, etc., when the user terminal receives a service data recommendation request for carrying the terminal identifier for the target service, The data recommendation device may sequentially send the at least one recommended service data set to the user terminal.
  • a recommended service data set may be sent every time a service data recommendation request is received; or, after a service data recommendation request is received, a recommended service data set may be sent at intervals, which is understandable. Yes, the recommended service data set sent this time is different from the recommended service data set sent by the history.
  • the step of receiving the service data recommendation request may be performed before the step of acquiring the timing service attribute corresponding to the terminal identifier of the user terminal and the real-time service attribute corresponding to the terminal identifier, that is, receiving the service data recommendation request, and then performing the recommended service. Selection of data Take the process.
  • the execution order of the above received service data recommendation request may be determined according to the service requirements of different target services.
  • the data recommendation device may add the service data recommendation request to a request queue constructed by using a distributed memory queue system, and it may be understood that a distributed memory queue system is adopted.
  • the constructed request queue has the characteristics of lightweight, high performance, easy to use, multi-queue, persistent, distributed fault tolerance and timeout control.
  • other memory queue systems can also be used to construct the request queue, for example: RabbitMQ, Kafka, Memcacheq or Fqueue.
  • the recommended service data is selected in the service data set in real time by acquiring the timing service attribute of the offline layer and the real-time service attribute of the real-time layer, and is pushed to the user terminal.
  • the process of realizing the recommended service data based on the service attributes provided by the offline layer and the real-time layer is realized, the update efficiency of the recommended service data is improved, the effect of the service data recommendation is improved, and the business data selection and real-time of the offline layer are shared.
  • the work of the service data recommendation of the layer improves the work efficiency of the service data recommendation.
  • the recommendation service data is selected based on the timed service attribute and the real-time service attribute, which satisfies different static and dynamic application scenarios and improves the user experience.
  • FIG. 3 is a schematic diagram of a system for providing a data recommendation method according to an embodiment of the present invention.
  • the system for performing the data recommendation method may include an offline layer, an offline data storage, an intermediate layer, and a real-time layer.
  • the offline layer, the offline data storage, the middle layer, and the real-time layer may be deployed in the same Different data blocks in the server can also be deployed in different servers.
  • the middle layer is configured to perform the method steps performed by the data recommendation device.
  • the pipeline data storage server is configured to store each terminal identifier at a preset time.
  • the first service operation performed on the target service in the segment is output to the pipeline server for processing, and the operation information corresponding to the first service operation is obtained, for example, generating a corresponding array matrix as input data of the recommended algorithm model
  • the recommended algorithm model may include a User-Based and Item-Based collaborative filtering algorithm model, a Content-Based recommendation algorithm model, an RBM model, an RNN model, etc., and the recommended algorithm model outputs a corresponding timing attribute label. It can be understood that Each terminal identifier has its own timing attribute label, and the offline layer can store the timing attribute label to the offline data storage.
  • the offline layer may also calculate other similar service data and the like with the existing service data under the terminal identifier, and the offline layer may also store similar service data identified by each terminal to the offline data storage.
  • the service data set corresponding to the target service may be stored, where the service data set may include at least one service data, and each of the at least one service data has a corresponding one of the service data.
  • the third tag attribute is further stored in the offline data storage, and the offline data storage further stores a timing attribute tag of each terminal identifier and similar service data.
  • the user terminal may send a service data recommendation request carrying the terminal identifier of the user terminal to a common gateway interface (CGI), and the CGI may transmit the service data recommendation request to the distributed memory.
  • CGI common gateway interface
  • the queue system the middle layer may pull the service data set, the third label attribute set, and the timing service attribute corresponding to the terminal identifier in the offline data storage, and the middle layer may also pull the terminal identifier from the real-time layer.
  • the intermediate layer may use the preset scoring model in the target service according to the timing service attribute and the real-time service attribute, and the third tag attribute set and the similar service data of the terminal identifier.
  • the recommended service data is selected in the corresponding service data set, and the preset scoring model may be a GBDT algorithm model.
  • the intermediate layer may further perform re-screening on the selected preset number of target service data with the highest score value based on a preset screening rule to obtain the recommended service data, and the screening Regulation Then, a screening rule set by the developer based on the service characteristic of the target service may be used, and the middle layer sends the recommended service data to the user terminal.
  • the real-time layer can obtain the second service operation corresponding to the terminal identifier of the user terminal in real time, and transmit it to the second service operation acquisition server, and calculate the real-time service attribute corresponding to each terminal identifier by using the confidence model, and the confidence model can be Including SGD algorithm model, FTRL algorithm model and so on.
  • the data recommendation device provided by the embodiment of the present invention will be described in detail below with reference to FIG. It should be noted that the communication connection device shown in FIG. 4 to FIG. 6 is configured to perform the method of the embodiment shown in FIG. 1 to FIG. 3, and for the convenience of description, only the embodiment of the present invention is shown. For the part, the technical details are not disclosed, please refer to the embodiment shown in Figs. 1 to 3 of the present invention.
  • FIG. 4 is a schematic structural diagram of a data recommendation device according to an embodiment of the present invention.
  • the data recommendation device 1 of the embodiment of the present invention may include: an attribute obtaining unit 11, a recommendation data selecting unit 12, and a data transmitting unit 13.
  • the attribute obtaining unit 11 is configured to acquire a timing service attribute corresponding to the terminal identifier of the user terminal and a real-time service attribute corresponding to the terminal identifier;
  • the attribute obtaining unit 11 may obtain the terminal identifier of the user terminal for the corresponding timing service attribute in the target service and the real-time service attribute corresponding to the terminal identifier, and the terminal identifier of the user terminal may be The unique identifier of the user terminal, or may be a user registration of the target service in advance, and use the user terminal to log in to the target service, such as an application account of the music application, online shopping. Application account number of the application.
  • the timing service attribute is a first service operation for the target service that is uploaded in the first preset time period according to the terminal identifier, and the first corresponding to the terminal identifier is calculated.
  • Tag attribute further, the timed service attribute may be based on The recorded terminal identifier identifies a first service attribute for the target service that is uploaded in the first preset time period, and uses the preset recommendation algorithm model to calculate the first label attribute corresponding to the terminal identifier.
  • the first label attribute is obtained by calculation of the offline layer, for example, for the music application, acquiring the first service operation for the song in the music application uploaded by the terminal identifier within one month
  • the operation includes downloading, skipping, deleting, and collecting operations.
  • the offline layer may adopt a preset recommendation algorithm model, such as a User-Based and Item-Based collaborative filtering algorithm model, a Content-Based recommendation algorithm model, an RBM model, An RNN model or the like, and calculating, according to the first operation under the terminal identifier, a first tag attribute corresponding to the terminal identifier, including a comprehensive music genre, a singer preference, an age preference, etc. within one month, and calculating the terminal Other similar music data, etc. of the music list under the logo;
  • a preset recommendation algorithm model such as a User-Based and Item-Based collaborative filtering algorithm model, a Content-Based recommendation algorithm model, an RBM model, An RNN model or the like
  • the real-time service attribute is a second label attribute corresponding to the terminal identifier calculated by the second service operation for the target service that is uploaded in real time according to the terminal identifier.
  • the real-time service attribute may be And the second label attribute corresponding to the terminal identifier calculated by using the preset confidence model, the second label attribute corresponding to the terminal identifier calculated by the terminal identifier in the second preset time period, A preset time period is greater than the second preset time period.
  • the second tag attribute is obtained by calculating the real-time layer, for example, for a relatively static application scenario such as a daily selection in a music application, acquiring the target that the terminal identifier is uploaded within 1 day.
  • the second business operation of the music data in the music application including downloading, skipping, deleting, collecting, and the like.
  • the real-time layer may adopt a preset confidence model, for example, an SGD algorithm model, an FTRL algorithm model, etc., and calculate a second label attribute corresponding to the terminal identifier according to the second service operation of the terminal identifier, including one day.
  • a preset confidence model for example, an SGD algorithm model, an FTRL algorithm model, etc.
  • the real-time service attribute may be a second service operation based on the recorded terminal identifier to the historical recommended service data, and the terminal identifier corresponding to the preset confidence model is used to calculate Real-time business attributes, the historical recommended business data is based
  • the last batch of recommended business data of the recommended business data for example, for a dynamic application scenario such as guessing your favorite in the music application, 15 music data can be selected as a batch of recommended business data at one time, and divided into 3 times.
  • the push operation is performed on each of the five items, so that the second service operation, such as downloading, skipping, deleting, and collecting, of the recommended service data (ie, historical recommended service data) uploaded by the terminal identifier can be obtained.
  • the real-time layer may also adopt a preset confidence model, and calculate a second label attribute corresponding to the terminal identifier according to the second service operation in the terminal identifier, including the second service operation of the recommended service data of the batch.
  • first service operation and the second service operation may be the same service operation, and the first and second divisions are only used to describe the difference between the timing and the real-time.
  • the recommendation data selection unit 12 is configured to select at least one recommended service data in the service data set corresponding to the target service according to the timing service attribute and the real-time service attribute;
  • the recommendation data selecting unit 12 may select at least one recommended service data in the service data set corresponding to the target service according to the timing service attribute and the real-time service attribute.
  • the service data set includes at least one service data, and each of the at least one service data has a third tag attribute corresponding to each of the service data, for example, for music.
  • the music data in the music library corresponds to a third tag attribute such as a music genre, a singer, a genre, etc.
  • the recommended data selecting unit 12 may obtain the third tag attribute in the at least one service data to satisfy the And timing the service attribute and the at least one target service data of the real-time service attribute, wherein the recommendation data extraction unit 12 may satisfy the third service attribute and all of the real-time service attribute in the service data set.
  • the service data is selected, and the at least one target service data represents all service data selected in the service data set according to the timing service attribute and the real-time service attribute.
  • the recommendation data selection unit 12 adopts a preset scoring model according to the timing service attribute and the real-time service attribute, and the preset scoring model may be a GBDT algorithm model. Acquiring a score value of each target service data in the at least one target service data, the recommendation data selection unit 12 acquiring at least one recommendation in the at least one target service data according to a size of a score value of each target service data Business data.
  • FIG. 5 is a schematic structural diagram of a recommendation data selection unit according to an embodiment of the present invention.
  • the recommendation data selection unit 12 may include:
  • the target data obtaining sub-unit 121 is configured to acquire, in the at least one service data, at least one target service data that the third tag attribute satisfies the timing service attribute and the real-time service attribute;
  • the service data set includes at least one service data, and each of the at least one service data has a third tag attribute corresponding to each service data, for example, for a music application, in a music library.
  • Each piece of music data corresponds to a third tag attribute such as a music genre, a singer, a genre, etc.
  • the target data obtaining sub-unit 121 may obtain, in the at least one service data, a third tag attribute that satisfies the timing service attribute and the Depicting at least one target service data of the real-time service attribute, the target data acquisition sub-unit 121 may select, in the service data set, all service data that the third tag attribute satisfies the timing service attribute and the real-time service attribute
  • the at least one target service data represents all service data selected in the service data set according to the timing service attribute and the real-time service attribute.
  • the target data obtaining sub-unit 121 may obtain, in the at least one service data, first candidate service data corresponding to the third tag attribute that matches the timing service attribute, for example, for Music application, there is A music data in the music data collection, and the third label attribute can be XX genre, YY singer, ZZ era, B music data exists, and the third label attribute can be OO genre, PP singer, QQ age, There is C music data, and the third tag attribute may be DD genre, EE singer, FF era, the timing service attribute is XX genre and QQ age, then the target data acquisition subunit 121 may be A sound The music data and the B music data are used as the first candidate business data.
  • the target data acquisition sub-unit 121 obtains second candidate service data corresponding to the third label attribute that matches the real-time service attribute in the at least one service data, and according to the above example, if the real-time layer detects the terminal identifier
  • the executed second business operation includes continuously collecting the music data of the EE singer, and the target data acquisition sub-unit 121 may use the C music data as the second candidate service data.
  • the target data acquisition sub-unit 121 determines the first candidate service data and the second candidate service data as at least one target service data.
  • the target data obtaining sub-unit 121 may further acquire, in the at least one service data, third candidate service data corresponding to the third tag attribute that matches the timing service attribute, for example:
  • third candidate service data corresponding to the third tag attribute that matches the timing service attribute
  • the target data acquisition sub-unit 121 may be A music The data, the B music data, and the C music data are used as the third candidate service data.
  • the target data acquisition sub-unit 121 deletes the fourth candidate service data corresponding to the third label attribute that matches the real-time service attribute in the third candidate service data.
  • the second service operation includes deleting the music data of the PP singer continuously, and the target data acquisition sub-unit 121 may use the B music data as the fourth candidate service data, and the target data acquisition sub-unit 121 deletes the The third candidate service data of the fourth candidate service data generates at least one target service data.
  • the value acquisition sub-unit 122 is configured to acquire a score value of each target service data in the at least one target service data according to the timing service attribute and the real-time service attribute and adopt a preset scoring model;
  • the value acquisition sub-unit 122 is based on the timing service attribute and the real-time service.
  • the preset scoring model may be a GBDT algorithm model, and obtain a scoring value of each target service data in the at least one target service data.
  • the value obtaining sub-unit 122 may The timing service attribute and the real-time service attribute are used as weight parameters of the preset scoring model, and a score value of each target service data in the at least one target service data is calculated.
  • the recommendation data acquisition sub-unit 123 is configured to acquire at least one recommended service data in the at least one target service data according to the size of the score value of each target service data;
  • the recommendation data acquisition sub-unit 123 acquires at least one recommended service data in the at least one target service data according to the size of the score value of each target service data.
  • the recommendation data acquisition sub-unit 123 may sort the at least one target service data according to the size of the score value of each target service data, and at least one of the sorted ones.
  • the target service data is selected from the target service data of the preset quantity and having the highest score value as the at least one recommended service data.
  • the recommendation data acquisition sub-unit 123 may further perform re-screening on the selected preset number of target service data with the highest score value based on a preset screening rule to obtain the at least one recommended service data, and the screening
  • the rule may be a screening rule set by the developer based on the business characteristics of the target service, for example, for a music application, the screening rule may specify that the selected preset number of music data with the highest score value cannot exist in the same singer. Tag attributes, etc.
  • the data sending unit 13 is configured to send the at least one recommended service data to the user terminal;
  • the data sending unit 13 may send the at least one recommended service data to the user terminal corresponding to the terminal identifier.
  • the data sending unit 13 may search for the terminal when receiving the service data recommendation request that is sent by the user terminal and carrying the terminal identifier. Identification pair The at least one recommended service data should be sent, and the at least one recommended service data is sent to the user terminal.
  • FIG. 6 is a schematic structural diagram of a data sending unit according to an embodiment of the present invention.
  • the data sending unit 13 may include:
  • the data sorting sub-unit 131 is configured to perform sorting processing on the at least one recommended service data according to a score value of each recommended service data in the at least one recommended service data;
  • the data set dividing sub-unit 132 is configured to divide the at least one recommended service data into at least one recommended service data set according to the at least one recommended service data after the sorting process;
  • the data set sending sub-unit 133 is configured to, when receiving the service data recommendation request that is sent by the user terminal for the target service and carrying the terminal identifier, send the at least one recommended service data set to the user terminal in sequence;
  • the data sorting sub-unit 131 may perform sorting processing on the at least one recommended service data according to a score value of each recommended service data in the at least one recommended service data, where the data set dividing sub-unit 132 is After the sorting process, the at least one recommended service data is divided into at least one recommended service data set, and each of the at least one recommended service data set includes a preset number of recommended service data, for example, For the music application, a total of 15 pieces of music data are selected as a batch of recommended service data, and are divided into 3 times, each time for 5 pushes, etc., when the receiving user terminal transmits the service data carrying the terminal identifier for the target service.
  • the data set transmission sub-unit 133 may sequentially send the at least one recommended service data set to the user terminal.
  • a recommended service data set may be sent every time a service data recommendation request is received; or, after a service data recommendation request is received, a recommended service data set may be sent at intervals, which is understandable. Yes, the recommended service data set sent this time is different from the recommended service data set sent by the history.
  • the step of receiving the service data recommendation request may also obtain the timing service attribute corresponding to the terminal identifier of the user terminal and the terminal. The step of identifying the corresponding real-time service attribute is performed before the step of receiving the service data recommendation request, and then the selection process of the recommended service data is performed.
  • the execution order of the above received service data recommendation request may be determined according to the service requirements of different target services.
  • the data recommendation device 1 may add the service data recommendation request to a request queue constructed by using a distributed memory queue system, and it may be understood that a distributed memory queue is adopted.
  • the system-built request queue has the characteristics of lightweight, high-performance, easy-to-use, multi-queue, persistent, distributed fault tolerance, and timeout control.
  • other memory queue systems can also be used to construct the request queue, for example: RabbitMQ , Kafka, Memcacheq or Fqueue.
  • the recommended service data is selected in the service data set in real time by acquiring the timing service attribute of the offline layer and the real-time service attribute of the real-time layer, and is pushed to the user terminal.
  • the process of realizing the recommended service data based on the service attributes provided by the offline layer and the real-time layer is realized, the update efficiency of the recommended service data is improved, the effect of the service data recommendation is improved, and the business data selection and real-time of the offline layer are shared.
  • the work of the service data recommendation of the layer improves the work efficiency of the service data recommendation.
  • the recommendation service data is selected based on the timed service attribute and the real-time service attribute, which satisfies different static and dynamic application scenarios and improves the user experience.
  • Each unit included in the data recommendation device in the embodiment of the present invention, and each subunit included in each unit may be implemented by a processor in the data recommendation device; wherein the functions implemented by the processor may of course be implemented by logic circuits.
  • the processor may be a central processing unit (CPU), a microprocessor (MPU), a digital signal processor (DSP), or a field programmable gate array (FPGA).
  • FIG. 7 is a schematic structural diagram of another data recommendation device according to an embodiment of the present invention.
  • the data recommendation device 1000 may include at least one processor 1001, such as a CPU, at least one network interface 1004, a user interface 1003, a memory 1005, and at least one communication bus 1002.
  • the communication bus 1002 is used to implement connection communication between these components.
  • the user interface 1003 may include a display and a keyboard.
  • the user interface 1003 may further include a standard wired interface and a wireless interface.
  • the network interface 1004 can include a standard wired interface, a wireless interface (such as a WI-FI interface).
  • the memory 1005 may be a high speed RAM memory or a non-volatile memory such as at least one disk memory.
  • the memory 1005 may also be at least one storage device located remotely from the aforementioned processor 1001. As shown in FIG. 7, an operating system, a network communication module, a user interface module, and a data recommendation application may be included in the memory 1005 as a computer storage medium.
  • the user interface 1003 is configured to provide an input interface for the user to acquire data input by the user;
  • the network interface 1004 is configured to be connected to the user terminal to acquire data of the user terminal; and
  • the processor The 1001 can be configured to invoke the data recommendation application stored in the memory 1005 and perform the following operations:
  • timing service attribute corresponding to the terminal identifier of the user terminal and a real-time service attribute corresponding to the terminal identifier, where the timing service attribute is a first service for the target service uploaded according to the terminal identifier in the first preset time period
  • first label attribute corresponding to the terminal identifier is calculated
  • real-time service attribute is a second service operation for the target service that is uploaded in real time according to the terminal identifier
  • the calculated terminal identifier corresponds to Second tag attribute
  • the processor 1001 is configured to acquire a terminal identifier pair of the user terminal.
  • the scheduled service attribute and the real-time service attribute corresponding to the terminal identifier are used, perform the following operations:
  • the first preset time period is greater than the second preset time period.
  • the processor 1001 when the processor 1001 performs the acquisition of the timing service attribute corresponding to the terminal identifier of the user terminal and the real-time service attribute corresponding to the terminal identifier, the processor 1001 performs the following operations:
  • the service data set includes at least one service data, and each of the at least one service data has a third tag attribute corresponding to each of the service data;
  • the processor 1001 performs the following operations when performing at least one recommended service data in the service data set corresponding to the target service according to the timing service attribute and the real-time service attribute:
  • the processor 1001 performs the following operations when performing, by acquiring, in the at least one service data, at least one target service data that the third tag attribute satisfies the timing service attribute and the real-time service attribute:
  • the first candidate service data and the second candidate service data are determined as at least one target service data.
  • the processor 1001 performs the following operations when performing, by acquiring, in the at least one service data, at least one target service data that the third tag attribute satisfies the timing service attribute and the real-time service attribute:
  • the processor 1001 performs the following operations when performing the sending of the at least one recommended service data to the user terminal:
  • each recommended service data set in the at least one recommended service data set includes a preset number of recommended services data
  • the at least one recommended service data set is sequentially sent to the user terminal.
  • the recommended service data is selected in the service data set in real time by acquiring the timing service attribute of the offline layer and the real-time service attribute of the real-time layer, and is pushed to the user terminal.
  • the process of realizing the recommended service data based on the service attributes provided by the offline layer and the real-time layer is realized, the update efficiency of the recommended service data is improved, the effect of the service data recommendation is improved, and the business data selection and real-time of the offline layer are shared.
  • the recommendation of the service data of the layer improves the work efficiency of the service data recommendation.
  • the recommendation service data is selected based on the timed service attribute and the real-time service attribute, which satisfies different static and dynamic application scenarios and improves the user experience.
  • the above data recommendation method is implemented in the form of a software function module and sold or used as a standalone product, it may also be stored in a computer readable storage medium.
  • the technical solution of the embodiments of the present invention may be embodied in the form of a software product in essence or in the form of a software product stored in a storage medium, including a plurality of instructions.
  • a computer device (which may be a personal computer, server, or network device, etc.) is caused to perform all or part of the methods described in various embodiments of the present invention.
  • the foregoing storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read only memory (ROM), a magnetic disk, or an optical disk.
  • program codes such as a USB flash drive, a mobile hard disk, a read only memory (ROM), a magnetic disk, or an optical disk.
  • an embodiment of the present invention further provides a computer storage medium, where the computer stores Computer executable instructions are stored in the medium for performing the data recommendation method in the embodiments of the present invention.
  • the storage medium may be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).
  • the time service attribute corresponding to the terminal identifier of the user terminal and the real-time service attribute corresponding to the terminal identifier are obtained, and the timed service attribute is the first service for the target service uploaded according to the terminal identifier in the first preset time period.
  • the operation, the first label attribute corresponding to the terminal identifier is calculated;
  • the real-time service attribute is a second service operation for the target service that is uploaded in real time according to the terminal identifier, and the second label attribute corresponding to the calculated terminal identifier;
  • the real-time service attribute selects at least one recommended service data in the service data set corresponding to the target service; and sends the at least one recommended service data to the user terminal.
  • the technical solution provided by the embodiment of the present invention can generate recommended service data in real time based on the service attributes provided by the offline layer and the real-time layer, improve the update efficiency of the recommended service data, and improve the effect of the service data recommendation.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Business, Economics & Management (AREA)
  • General Physics & Mathematics (AREA)
  • Accounting & Taxation (AREA)
  • Finance (AREA)
  • Software Systems (AREA)
  • Development Economics (AREA)
  • Strategic Management (AREA)
  • Data Mining & Analysis (AREA)
  • Mathematical Physics (AREA)
  • General Engineering & Computer Science (AREA)
  • Computing Systems (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Signal Processing (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Computational Linguistics (AREA)
  • Marketing (AREA)
  • General Business, Economics & Management (AREA)
  • Economics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Game Theory and Decision Science (AREA)
  • Biomedical Technology (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biophysics (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Medical Informatics (AREA)
  • Probability & Statistics with Applications (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Information Transfer Between Computers (AREA)

Abstract

本发明实施例公开一种数据推荐方法及其设备、存储介质,其中方法包括:获取用户终端的终端标识对应的定时业务属性以及终端标识对应的实时业务属性,定时业务属性为根据终端标识在第一预设时间段内上传的针对目标业务的第一业务操作,所计算得到终端标识对应的第一标签属性;实时业务属性为根据终端标识实时上传的针对目标业务的第二业务操作,所计算得到的终端标识对应的第二标签属性;根据定时业务属性和实时业务属性在目标业务对应的业务数据集合中选取至少一个推荐业务数据;将至少一个推荐业务数据发送至用户终端。

Description

一种数据推荐方法及其设备、存储介质
本专利申请要求2016年05月12日提交的中国专利申请号为201610316826.8,申请人为腾讯科技(深圳)有限公司,发明名称为“一种数据推荐方法及其设备”的优先权,该申请的全文以引用的方式并入本申请中。
技术领域
本发明涉及计算机技术领域,尤其涉及一种数据推荐方法及其设备、存储介质。
背景技术
随着计算机技术不断的开发和完善,在使用终端应用的过程中,后台服务器除了保证终端应用的正常运行外,还会对相关业务数据进行推荐,例如:音乐应用中推荐每日精选歌曲、购物应用中推荐热卖商品等,提升了用户体验。
现有的推荐架构是由离线层和实时层两部分组成,离线层负责统计一段时间内用户针对当前业务的行为操作,从而计算每个用户的标签属性、相似数据等定时业务属性,同时还需要负责对推荐业务数据的选取;实时层则负责对推荐业务数据的再筛选,并将筛选后的业务数据输出至用户终端中,同时还需要根据用户针对当前业务的实时行为操作进行反馈学习,以生成每个用户的实时的标签属性等实时业务属性。由于离线层计算负担较大,容易影响推荐的业务数据的更新效率,并且由于离线层产生推荐的业务数据的周期较长,而实时层实时反馈学习所得到的业务属性往往需要等待离线层在下一次选取推荐的业务数据时才能生效,因此导致无法实时调整所推荐的业务数据,进而影响了业务数据推荐的效果。
发明内容
本发明实施例提供一种数据推荐方法及其设备、存储介质,可以基于离线层和实时层提供的业务属性,实时生成推荐业务数据,提高推荐的业务数据的更新效率,提升业务数据推荐的效果。
本发明实施例提供了一种数据推荐方法,可包括:
获取用户终端的终端标识对应的定时业务属性以及所述终端标识对应的实时业务属性,所述定时业务属性为根据所述终端标识在第一预设时间段内上传的针对目标业务的第一业务操作,所计算得到所述终端标识对应的第一标签属性;所述实时业务属性为根据所述终端标识实时上传的针对所述目标业务的第二业务操作,所计算得到的所述终端标识对应的第二标签属性;
根据所述定时业务属性和所述实时业务属性在所述目标业务对应的业务数据集合中选取至少一个推荐业务数据;
将所述至少一个推荐业务数据发送至所述用户终端。
本发明实施例提供了一种数据推荐设备,可包括:
属性获取单元,配置为获取用户终端的终端标识对应的定时业务属性以及所述终端标识对应的实时业务属性,所述定时业务属性为根据所述终端标识在第一预设时间段内上传的针对目标业务的第一业务操作,所计算得到所述终端标识对应的第一标签属性;所述实时业务属性为根据所述终端标识实时上传的针对所述目标业务的第二业务操作,所计算得到的所述终端标识对应的第二标签属性;
推荐数据选取单元,配置为根据所述定时业务属性和所述实时业务属性在所述目标业务对应的业务数据集合中选取至少一个推荐业务数据;
数据发送单元,配置为将所述至少一个推荐业务数据发送至所述用户终端。
本发明实施例提供一种计算机存储介质,所述计算机存储介质中存储有计算机可执行指令,该计算机可执行指令用于执行本发明实施例提供的数据推荐方法。
本发明实施例提供一种数据推荐设备,包括:
存储介质,配置为存储可执行指令;
处理器,配置为执行存储的可执行指令,所述可执行指令用于执行下面的步骤:
获取用户终端的终端标识对应的定时业务属性以及所述终端标识对应的实时业务属性,所述定时业务属性为根据所述终端标识在第一预设时间段内上传的针对目标业务的第一业务操作,所计算得到所述终端标识对应的第一标签属性;所述实时业务属性为根据所述终端标识实时上传的针对所述目标业务的第二业务操作,所计算得到的所述终端标识对应的第二标签属性;
根据所述定时业务属性和所述实时业务属性在所述目标业务对应的业务数据集合中选取至少一个推荐业务数据;
将所述至少一个推荐业务数据发送至所述用户终端。
在本发明实施例中,通过获取离线层的定时业务属性以及实时层的实时业务属性,在业务数据集合中实时选取推荐业务数据,并推送至用户终端。实现了基于离线层和实时层提供的业务属性,实时生成推荐业务数据的过程,提高了推荐的业务数据的更新效率,提升了业务数据推荐的效果,同时分担了离线层的业务数据选取以及实时层的业务数据推荐的工作,进而提升了业务数据推荐的工作效率。
附图说明
为了更清楚地说明本发明实施例或现有技术中的技术方案,下面将对实施例或现有技术描述中所需要使用的附图作简单地介绍,显而易见地, 下面描述中的附图仅仅是本发明的一些实施例,对于本领域普通技术人员来讲,在不付出创造性劳动的前提下,还可以根据这些附图获得其他的附图。
图1是本发明实施例提供的一种数据推荐方法的流程示意图;
图2是本发明实施例提供的另一种数据推荐方法的流程示意图;
图3是本发明实施例提供的一种数据推荐方法的***举例示意图;
图4是本发明实施例提供的一种数据推荐设备的结构示意图;
图5是本发明实施例提供的推荐数据选取单元的结构示意图;
图6是本发明实施例提供的数据发送单元的结构示意图;
图7是本发明实施例提供的另一种数据推荐设备的结构示意图。
具体实施方式
下面将结合本发明实施例中的附图,对本发明实施例中的技术方案进行清楚、完整地描述,显然,所描述的实施例仅仅是本发明一部分实施例,而不是全部的实施例。基于本发明中的实施例,本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例,都属于本发明保护的范围。
本发明实施例提供的数据推荐方法可以应用于音乐数据推荐的场景,例如:数据推荐设备获取用户终端的终端标识在音乐应用中对应的定时音乐标签属性以及所述终端标识对应的实时音乐标签属性,所述数据推荐设备根据所述定时音乐标签属性和所述实时音乐标签属性在所述音乐数据集合中选取至少一个推荐音乐数据,所述数据推荐设备将所述至少一个推荐音乐数据发送至所述用户终端的场景;还可以应用于商品数据推荐的场景,例如:数据推荐设备获取用户终端的终端标识在线上购物应用中对应的定时商品标签属性以及所述终端标识对应的实时商品标签属性,所述数据推荐设备根据所述定时商品标签属性和所述实时商品标签属性在所述线上购 物应用对应的商品数据集合中选取至少一个推荐商品数据,所述数据推荐设备将所述至少一个推荐商品数据发送至所述用户终端的场景等。实现了基于离线层和实时层提供的业务属性,实时生成推荐业务数据的过程,提高了推荐的业务数据的更新效率,提升了业务数据推荐的效果,同时分担了离线层的业务数据选取以及实时层的业务数据推荐的工作,进而提升了业务数据推荐的工作效率。
本发明实施例涉及的数据推荐设备可以为目标业务对应的后台服务设备,配置为根据用户终端的终端标识对应的不同维度的标签属性,生成对应的推荐业务数据并推送至用户终端;所述用户终端可以包括平板电脑、智能手机、掌上电脑以及移动互联网设备(MID)等具备支持目标业务等功能在内的终端设备;所述目标业务可以为提供业务需求的应用,例如:音乐应用、电影应用、线上购物应用等。
下面将结合附图1和附图2,对本发明实施例提供的数据推荐方法进行详细介绍。
请参见图1,为本发明实施例提供了一种数据推荐方法的流程示意图。如图1所示,本发明实施例的所述方法可以包括以下步骤S101-步骤S103。
S101,获取用户终端的终端标识对应的定时业务属性以及所述终端标识对应的实时业务属性;
在本发明的其他实施例中,数据推荐设备可以获取用户终端的终端标识针对在目标业务中对应的定时业务属性以及所述终端标识对应的实时业务属性,需要说明的是,所述用户终端的终端标识可以为所述用户终端的唯一识别码,或者可以为预先对所述目标业务进行用户注册,并使用所述用户终端登录所述目标业务的唯一业务标识,例如:音乐应用的应用账号、线上购物应用的应用账号等。
在本发明实施例中,所述定时业务属性为根据所述终端标识在第一预 设时间段内上传的针对所述目标业务的第一业务操作,所计算得到所述终端标识对应的第一标签属性,进一步的,所述定时业务属性可以为基于记录的所述终端标识在第一预设时间段内上传的针对目标业务的第一业务操作,并采用预设的推荐算法模型计算得到的所述终端标识对应的第一标签属性。在本发明的其他实施例中,第一标签属性为离线层通过计算得到的,例如:针对音乐应用,获取所述终端标识在1个月内上传的针对音乐应用中的歌曲的第一业务操作,包括下载、跳过、删除、收藏等操作,所述离线层可以采用预设的推荐算法模型,例如基于用户(User-Based)和基于类别(Item-Based)的协同过滤算法模型、基于内容(Content-Based)推荐算法模型、受限玻尔兹曼机(Restricted Boltzmann Machine,RBM)模型、循环神经网络(Recurrent Neural Networks,RNN)模型等,并根据该终端标识下的第一操作计算出该终端标识对应的第一标签属性,包括1个月内的综合的音乐流派、歌手偏好、年代偏好等,还可以计算与该终端标识下的音乐列表的其它相似的音乐数据等;
所述实时业务属性为根据所述终端标识实时上传的针对所述目标业务的第二业务操作,所计算得到的所述终端标识对应的第二标签属性,进一步的,所述实时业务属性可以为基于记录的所述终端标识在第二预设时间段内上传的针对目标业务的第二业务操作,并采用预设的置信模型计算得到的所述终端标识对应的第二标签属性,所述第一预设时间段大于所述第二预设时间段。在本发明的其他实施例中,第二标签属性为实时层通过计算得到的,例如:针对音乐应用中的每日精选等相对静态的应用场景,获取所述终端标识在1天内上传的针对音乐应用中的音乐数据的第二业务操作,包括下载、跳过、删除、收藏等操作。所述实时层可以采用预设的置信模型,例如:随机梯度下降(Stochastic Gradient Descent,SGD)算法模型、遵循规范的领导者(FTRL,Follow The Regularized Leader)算法模型 等,并根据该终端标识下的第二业务操作计算出该终端标识对应的第二标签属性,包括1天内的综合的音乐流派、歌手偏好、年代偏好等。
在本发明的其他实施例中,所述实时业务属性可以为基于记录的所述终端标识对历史推荐业务数据的第二业务操作,并采用预设的置信模型计算得到的所述终端标识对应的实时业务属性,所述历史推荐业务数据为基于本次推荐业务数据的上一批推荐业务数据,例如:针对音乐应用中的猜你喜欢等动态的应用场景,可以一次性选取15首音乐数据作为一批推荐业务数据,并分3次,每次5首进行推送,因此可以获取所述终端标识上传的对本批推荐业务数据(即历史推荐业务数据)的第二业务操作,包括下载、跳过、删除、收藏等操作。所述实时层同样可以采用预设的置信模型,并根据该终端标识下的第二业务操作计算出该终端标识对应的第二标签属性,包括对本批推荐业务数据的第二业务操作所体现的音乐流派、歌手偏好、年代偏好等。
需要说明的是,所述第一业务操作和所述第二业务操作可以为相同的业务操作,采用第一和第二进行区分仅为了说明定时和实时的区别。
S102,根据所述定时业务属性和所述实时业务属性在所述目标业务对应的业务数据集合中选取至少一个推荐业务数据;
在本发明的其他实施例中,所述数据推荐设备可以根据所述定时业务属性和所述实时业务属性在所述目标业务对应的业务数据集合中选取至少一个推荐业务数据。在本发明的其他实施例中,所述业务数据集合包括至少一个业务数据,所述至少一个业务数据中每个业务数据均存在所述每个业务数据对应的第三标签属性,例如:针对音乐应用,音乐库中的每一首音乐数据都对应有音乐流派、歌手、年代等第三标签属性,所述数据推荐设备可以在所述至少一个业务数据中获取第三标签属性满足所述定时业务属性和所述实时业务属性的至少一个目标业务数据,所述数据推荐设备可 以在所述业务数据集合中将所述第三标签属性满足定时业务属性和所述实时业务属性的所有业务数据选取出来,所述至少一个目标业务数据表示根据所述定时业务属性和所述实时业务属性在所述业务数据集合中选取的所有业务数据。所述数据推荐设备根据定时业务属性和所述实时业务属性并采用预设评分模型,所述预设评分模型可以为迭代决策树(Gradient Boosting Decision Tree,GBDT)算法模型,获取所述至少一个目标业务数据中每个目标业务数据的评分数值,所述数据推荐设备按照每个目标业务数据的评分数值的大小,在所述至少一个目标业务数据中获取至少一个推荐业务数据。
S103,将所述至少一个推荐业务数据发送至所述用户终端;
在本发明的其他实施例中,所述数据推荐设备可以将所述至少一个推荐业务数据发送至所述终端标识对应的所述用户终端。在本发明的其他实施例中,针对上述相对静态的应用场景,所述数据推荐设备可以在接收到所述用户终端发送的携带有所述终端标识的业务数据推荐请求时,查找所述终端标识对应的所述至少一个推荐业务数据,并将所述至少一个推荐业务数据发送至所述用户终端;针对上述动态的应用场景,所述数据推荐设备可以依据所述至少一个推荐业务数据中每个推荐业务数据的评分数值,对所述至少一个推荐业务数据进行排序处理,并在排序处理后,对所述至少一个推荐业务数据进行划分处理,划分为至少一个推荐业务数据集合,所述至少一个推荐业务数据集合中每个推荐业务数据集合包含预设数量的推荐业务数据,例如:针对音乐应用,一共选取15首音乐数据作为一批推荐业务数据,并分3次,每次5首进行推送等,当接收用户终端针对目标业务发送的携带有所述终端标识的业务数据推荐请求时,所述数据推荐设备可以依次将所述至少一个推荐业务数据集合发送至所述用户终端。在本发明的其他实施例中,可以每接收一次业务数据推荐请求,发送一个推荐 业务数据集合;或者,可以接收一次业务数据推荐请求后,每隔一段时间发送一个推荐业务数据集合,可以理解的是,本次发送的推荐业务数据集合区别于历史发送的推荐业务数据集合。当然,接收业务数据推荐请求的步骤也可以在获取用户终端的终端标识对应的定时业务属性以及所述终端标识对应的实时业务属性的步骤之前执行,即接收到业务数据推荐请求,才执行推荐业务数据的选取过程。以上接收业务数据推荐请求的执行顺序可以根据不同的目标业务的业务需求进行决定。
在本发明实施例中,通过获取离线层的定时业务属性以及实时层的实时业务属性,在业务数据集合中实时选取推荐业务数据,并推送至用户终端。实现了基于离线层和实时层提供的业务属性,实时生成推荐业务数据的过程,提高了推荐的业务数据的更新效率,提升了业务数据推荐的效果,同时分担了离线层的业务数据选取以及实时层的业务数据推荐的工作,进而提升了业务数据推荐的工作效率。
请参见图2,为本发明实施例提供了另一种数据推荐方法的流程示意图。如图2所示,本发明实施例的所述方法可以包括以下步骤S201至步骤S207。
S201,获取用户终端的终端标识对应的定时业务属性以及所述终端标识对应的实时业务属性;
在本发明的其他实施例中,数据推荐设备可以获取用户终端的终端标识针对在目标业务中对应的定时业务属性以及所述终端标识对应的实时业务属性,需要说明的是,所述用户终端的终端标识可以为所述用户终端的唯一识别码,或者可以为预先对所述目标业务进行用户注册,并使用所述用户终端登录所述目标业务的唯一业务标识,例如:音乐应用的应用账号、线上购物应用的应用账号等。
在本发明实施例中,所述定时业务属性为根据所述终端标识在第一预 设时间段内上传的针对所述目标业务的第一业务操作,所计算得到所述终端标识对应的第一标签属性,进一步的,所述定时业务属性可以为基于记录的所述终端标识在第一预设时间段内上传的针对目标业务的第一业务操作,并采用预设的推荐算法模型计算得到的所述终端标识对应的第一标签属性。在本发明的其他实施例中,第一标签属性为离线层通过计算得到的,例如:针对音乐应用,获取所述终端标识在1个月内上传的针对音乐应用中的歌曲的第一业务操作,包括下载、跳过、删除、收藏等操作,所述离线层可以采用预设的推荐算法模型,例如User-Based和Item-Based的协同过滤算法模型、Content-Based推荐算法模型、RBM模型、RNN模型等,并根据该终端标识下的第一操作计算出该终端标识对应的第一标签属性,包括1个月内的综合的音乐流派、歌手偏好、年代偏好等,还可以计算与该终端标识下的音乐列表的其它相似的音乐数据等;
所述实时业务属性为根据所述终端标识实时上传的针对所述目标业务的第二业务操作,所计算得到的所述终端标识对应的第二标签属性,进一步的,所述实时业务属性可以为基于记录的所述终端标识在第二预设时间段内上传的针对目标业务的第二业务操作,并采用预设的置信模型计算得到的所述终端标识对应的第二标签属性,所述第一预设时间段大于所述第二预设时间段。在本发明的其他实施例中,第二标签属性为实时层通过计算得到的,例如:针对音乐应用中的每日精选等相对静态的应用场景,获取所述终端标识在1天内上传的针对音乐应用中的音乐数据的第二业务操作,包括下载、跳过、删除、收藏等操作。所述实时层可以采用预设的置信模型,例如:SGD算法模型、FTRL算法模型等,并根据该终端标识下的第二业务操作计算出该终端标识对应的第二标签属性,包括1天内的综合的音乐流派、歌手偏好、年代偏好等。
在本发明的其他实施例中,所述实时业务属性可以为基于记录的所述 终端标识对历史推荐业务数据的第二业务操作,并采用预设的置信模型计算得到的所述终端标识对应的实时业务属性,所述历史推荐业务数据为基于本次推荐业务数据的上一批推荐业务数据,例如:针对音乐应用中的猜你喜欢等动态的应用场景,可以一次性选取15首音乐数据作为一批推荐业务数据,并分3次,每次5首进行推送,因此可以获取所述终端标识上传的对本批推荐业务数据(即历史推荐业务数据)的第二业务操作,包括下载、跳过、删除、收藏等操作。所述实时层同样可以采用预设的置信模型,并根据该终端标识下的第二业务操作计算出该终端标识对应的第二标签属性,包括对本批推荐业务数据的第二业务操作所体现的音乐流派、歌手偏好、年代偏好等。
需要说明的是,所述第一业务操作和所述第二业务操作可以为相同的业务操作,采用第一和第二进行区分仅为了说明定时和实时的区别。
S202,在所述至少一个业务数据中获取第三标签属性满足所述定时业务属性和所述实时业务属性的至少一个目标业务数据;
在本发明的其他实施例中,所述业务数据集合包括至少一个业务数据,所述至少一个业务数据中每个业务数据均存在所述每个业务数据对应的第三标签属性,例如:针对音乐应用,音乐库中的每一首音乐数据都对应有音乐流派、歌手、年代等第三标签属性,所述数据推荐设备可以在所述至少一个业务数据中获取第三标签属性满足所述定时业务属性和所述实时业务属性的至少一个目标业务数据,所述数据推荐设备可以在所述业务数据集合中将第三标签属性满足所述定时业务属性和所述实时业务属性的所有业务数据选取出来,所述至少一个目标业务数据表示根据所述定时业务属性和所述实时业务属性在所述业务数据集合中选取的所有业务数据。
在本发明的其他实施例中,所述数据推荐设备可以在所述至少一个业务数据中获取与所述定时业务属性匹配的第三标签属性对应的第一候选业 务数据,例如:针对音乐应用,在音乐数据集合中存在A音乐数据,其第三标签属性可以为XX流派、YY歌手、ZZ年代,存在B音乐数据,其第三标签属性可以为OO流派、PP歌手、QQ年代,存在C音乐数据,其第三标签属性可以为DD流派、EE歌手、FF年代,所述定时业务属性为XX流派和QQ年代,则所述数据推荐设备可以将A音乐数据和B音乐数据作为第一候选业务数据。所述数据推荐设备在所述至少一个业务数据中获取与所述实时业务属性匹配的第三标签属性对应的第二候选业务数据,依据上述举例,若实时层检测到所述终端标识执行的第二业务操作包括连续对EE歌手的音乐数据进行收藏,则所述数据推荐设备可以将C音乐数据作为第二候选业务数据。所述数据推荐设备将所述第一候选业务数据和所述第二候选业务数据确定为至少一个目标业务数据。
在本发明的其他实施例中,所述数据推荐设备还可以在所述至少一个业务数据中获取与所述定时业务属性匹配的第三标签属性对应的第三候选业务数据,例如:针对音乐应用,在音乐数据集合中存在A音乐数据,其第三标签属性可以为XX流派、YY歌手、ZZ年代,存在B音乐数据,其第三标签属性可以为OO流派、PP歌手、QQ年代,存在C音乐数据,其第三标签属性可以为DD流派、EE歌手、FF年代,所述定时业务属性为XX流派、PP歌手和FF年代,则所述数据推荐设备可以将A音乐数据、B音乐数据和C音乐数据作为第三候选业务数据。所述数据推荐设备在第三候选业务数据中删除与所述实时业务属性匹配的第三标签属性对应的第四候选业务数据,依据上述举例,若实时层检测到所述终端标识执行的第二业务操作包括连续对PP歌手的音乐数据进行删除,则所述数据推荐设备可以将B音乐数据作为第四候选业务数据,所述数据推荐设备根据删除了所述第四候选业务数据的第三候选业务数据生成至少一个目标业务数据。
S203,根据定时业务属性和所述实时业务属性并采用预设评分模型, 获取所述至少一个目标业务数据中每个目标业务数据的评分数值;
在本发明的其他实施例中,所述数据推荐设备根据定时业务属性和所述实时业务属性并采用预设评分模型,所述预设评分模型可以为GBDT算法模型,获取所述至少一个目标业务数据中每个目标业务数据的评分数值,进一步的,所述数据推荐设备可以将所述定时业务属性和所述实时业务属性作为所述预设评分模型的权重参数,计算所述至少一个目标业务数据中每个目标业务数据的评分数值。
S204,按照每个目标业务数据的评分数值的大小,在所述至少一个目标业务数据中获取至少一个推荐业务数据;
在本发明的其他实施例中,所述数据推荐设备按照每个目标业务数据的评分数值的大小,在所述至少一个目标业务数据中获取至少一个推荐业务数据。在本发明的其他实施例中,所述数据推荐设备可以按照所述每个目标业务数据的评分数值的大小,对所述至少一个目标业务数据进行排序,并在排序后的至少一个目标业务数据中选取预设数量的且评分数值最高的目标业务数据作为至少一个推荐业务数据。
进一步的,所述数据推荐设备还可以基于预设的筛选规则对选取的预设数量的且评分数值最高的目标业务数据进行再筛选,得到所述至少一个推荐业务数据,所述筛选规则可以为开发人员基于所述目标业务的业务特性设定的筛选规则,例如:针对音乐应用,所述筛选规则可以规定选取的预设数量的且评分数值最高的音乐数据中不能存在相同歌手的标签属性等。
在本发明实施例中,所述数据推荐设备可以将所述至少一个推荐业务数据发送至所述终端标识对应的所述用户终端。在本发明的其他实施例中,针对上述相对静态的应用场景,所述数据推荐设备可以在接收到所述用户终端发送的携带有所述终端标识的业务数据推荐请求时,查找所述终端标 识对应的所述至少一个推荐业务数据,并将所述至少一个推荐业务数据发送至所述用户终端;而针对上述动态的应用场景,可以参见本发明实施例的以下步骤S205-S207:
S205,按照所述至少一个推荐业务数据中每个推荐业务数据的评分数值,对所述至少一个推荐业务数据进行排序处理;
S206,根据排序处理后的至少一个推荐业务数据,将所述至少一个推荐业务数据划分为至少一个推荐业务数据集合;
S207,当接收用户终端针对目标业务发送的携带有所述终端标识的业务数据推荐请求时,依次将所述至少一个推荐业务数据集合发送至所述用户终端;
在本发明的其他实施例中,所述数据推荐设备可以依据所述至少一个推荐业务数据中每个推荐业务数据的评分数值,对所述至少一个推荐业务数据进行排序处理,并在排序处理后,对所述至少一个推荐业务数据进行划分处理,划分为至少一个推荐业务数据集合,所述至少一个推荐业务数据集合中每个推荐业务数据集合包含预设数量的推荐业务数据,例如:针对音乐应用,一共选取15首音乐数据作为一批推荐业务数据,并分3次,每次5首进行推送等,当接收用户终端针对目标业务发送的携带有所述终端标识的业务数据推荐请求时,所述数据推荐设备可以依次将所述至少一个推荐业务数据集合发送至所述用户终端。在本发明的其他实施例中,可以每接收一次业务数据推荐请求,发送一个推荐业务数据集合;或者,可以接收一次业务数据推荐请求后,每隔一段时间发送一个推荐业务数据集合,可以理解的是,本次发送的推荐业务数据集合区别于历史发送的推荐业务数据集合。当然,接收业务数据推荐请求的步骤也可以在获取用户终端的终端标识对应的定时业务属性以及所述终端标识对应的实时业务属性的步骤之前执行,即接收到业务数据推荐请求,才执行推荐业务数据的选 取过程。以上接收业务数据推荐请求的执行顺序可以根据不同的目标业务的业务需求进行决定。
在本发明的其他实施例中,所述数据推荐设备可以将所述业务数据推荐请求添加至采用分布式内存队列***所构建的请求队列中等待处理,可以理解的是,采用分布式内存队列***构建的请求队列具有轻量级、高性能、易使用、多队列、持久化、分布式容错以及超时控制等特性,当然,也可以采用其它的内存队列***进行请求队列的构建,例如:RabbitMQ、Kafka、Memcacheq或Fqueue等。
在本发明实施例中,通过获取离线层的定时业务属性以及实时层的实时业务属性,在业务数据集合中实时选取推荐业务数据,并推送至用户终端。实现了基于离线层和实时层提供的业务属性,实时生成推荐业务数据的过程,提高了推荐的业务数据的更新效率,提升了业务数据推荐的效果,同时分担了离线层的业务数据选取以及实时层的业务数据推荐的工作,进而提升了业务数据推荐的工作效率;通过基于定时业务属性以及实时业务属性实现推荐业务数据的选取,满足了不同的静态以及动态的应用场景,提升了用户体验;通过采用分布式内存队列***构建请求队列,可以使得请求队列具备轻量级、高性能、易使用、多队列、持久化、分布式容错以及超时控制等特性。
请参见图3,为本发明实施例提供了一种数据推荐方法的***举例示意图。如图3所示,执行数据推荐方法的***中可以包括离线层、离线数据存储、中间层以及实时层,可以理解的是,离线层、离线数据存储、中间层以及实时层可以部署在同一个服务器中的不同数据块中,也可以在不同的服务器中进行部署。其中,中间层配置为执行上述数据推荐设备所执行的方法步骤。
针对离线层,流水数据存储服务器配置为存储各终端标识在预设时间 段内针对目标业务执行的第一业务操作,并分别输出至流水服务器中进行处理,得到第一业务操作对应的操作信息,例如:生成相应的数组矩阵,作为推荐算法模型的输入数据,所述推荐算法模型可以包括User-Based和Item-Based的协同过滤算法模型、Content-Based推荐算法模型、RBM模型、RNN模型等,由所述推荐算法模型输出对应的定时属性标签,可以理解的是,每个终端标识均具备各自的定时属性标签,所述离线层可以将定时属性标签存放至离线数据存储。所述离线层还可以计算与该终端标识下的已有的业务数据的其它相似的业务数据等,所述离线层同样可以将每个终端标识的相似的业务数据存放至离线数据存储。
针对离线数据存储,可以存储由目标业务对应的业务数据集合,所述业务数据集合可以包括至少一个业务数据,所述至少一个业务数据中每个业务数据均存在所述每个业务数据对应的第三标签属性,因此所述离线数据存储中还存储有第三标签属性集合,所述离线数据存储中还存储有每个终端标识各自的定时属性标签以及相似的业务数据。
针对中间层,用户终端可以向公共网关接口(Common Gateway Interface,CGI)发送携带有所述用户终端的终端标识的业务数据推荐请求,所述CGI可以将所述业务数据推荐请求传输至分布式内存队列***,中间层可以在离线数据存储中拉取业务数据集合、第三标签属性集合以及所述终端标识对应的定时业务属性,所述中间层还可以从实时层中拉取所述终端标识的实时业务属性,所述中间层可以根据所述定时业务属性和所述实时业务属性、以及第三标签属性集合和所述终端标识的相似的业务数据,并采用预设评分模型在所述目标业务对应的业务数据集合中选取推荐业务数据,所述预设评分模型可以为GBDT算法模型。在本发明的其他实施例中,所述中间层还可以基于预设的筛选规则对选取的预设数量的且评分数值最高的目标业务数据进行再筛选,得到所述推荐业务数据,所述筛选规 则可以为开发人员基于所述目标业务的业务特性设定的筛选规则,所述中间层将所述推荐业务数据发送至所述用户终端。
针对实时层,实时层可以实时获取用户终端的终端标识对应的第二业务操作,传输至第二业务操作获取服务器,通过置信模型计算得到每个终端标识对应的实时业务属性,所述置信模型可以包括SGD算法模型、FTRL算法模型等。
下面将结合附图4-附图6,对本发明实施例提供的数据推荐设备进行详细介绍。需要说明的是,附图4-附图6所示的通信连接设备,配置为执行本发明图1-图3所示实施例的方法,为了便于说明,仅示出了与本发明实施例相关的部分,技术细节未揭示的,请参照本发明图1-图3所示的实施例。
请参见图4,为本发明实施例提供了一种数据推荐设备的结构示意图。如图4所示,本发明实施例的所述数据推荐设备1可以包括:属性获取单元11、推荐数据选取单元12和数据发送单元13。
属性获取单元11,配置为获取用户终端的终端标识对应的定时业务属性以及所述终端标识对应的实时业务属性;
实现中,所述属性获取单元11可以获取用户终端的终端标识针对在目标业务中对应的定时业务属性以及所述终端标识对应的实时业务属性,需要说明的是,所述用户终端的终端标识可以为所述用户终端的唯一识别码,或者可以为预先对所述目标业务进行用户注册,并使用所述用户终端登录所述目标业务的唯一业务标识,例如:音乐应用的应用账号、线上购物应用的应用账号等。
在本发明实施例中,所述定时业务属性为根据所述终端标识在第一预设时间段内上传的针对所述目标业务的第一业务操作,所计算得到所述终端标识对应的第一标签属性,进一步的,所述定时业务属性可以为基于记 录的所述终端标识在第一预设时间段内上传的针对目标业务的第一业务操作,并采用预设的推荐算法模型计算得到的所述终端标识对应的第一标签属性。在本发明的其他实施例中,第一标签属性为离线层通过计算得到的,例如:针对音乐应用,获取所述终端标识在1个月内上传的针对音乐应用中的歌曲的第一业务操作,包括下载、跳过、删除、收藏等操作,所述离线层可以采用预设的推荐算法模型,例如User-Based和Item-Based的协同过滤算法模型、Content-Based推荐算法模型、RBM模型、RNN模型等,并根据该终端标识下的第一操作计算出该终端标识对应的第一标签属性,包括1个月内的综合的音乐流派、歌手偏好、年代偏好等,还可以计算与该终端标识下的音乐列表的其它相似的音乐数据等;
所述实时业务属性为根据所述终端标识实时上传的针对所述目标业务的第二业务操作,所计算得到的所述终端标识对应的第二标签属性,进一步的,所述实时业务属性可以为基于记录的所述终端标识在第二预设时间段内上传的针对目标业务的第二业务操作,并采用预设的置信模型计算得到的所述终端标识对应的第二标签属性,所述第一预设时间段大于所述第二预设时间段。在本发明的其他实施例中,第二标签属性为实时层通过计算得到的,例如:针对音乐应用中的每日精选等相对静态的应用场景,获取所述终端标识在1天内上传的针对音乐应用中的音乐数据的第二业务操作,包括下载、跳过、删除、收藏等操作。所述实时层可以采用预设的置信模型,例如:SGD算法模型、FTRL算法模型等,并根据该终端标识下的第二业务操作计算出该终端标识对应的第二标签属性,包括1天内的综合的音乐流派、歌手偏好、年代偏好等。
在本发明的其他实施例中,所述实时业务属性可以为基于记录的所述终端标识对历史推荐业务数据的第二业务操作,并采用预设的置信模型计算得到的所述终端标识对应的实时业务属性,所述历史推荐业务数据为基 于本次推荐业务数据的上一批推荐业务数据,例如:针对音乐应用中的猜你喜欢等动态的应用场景,可以一次性选取15首音乐数据作为一批推荐业务数据,并分3次,每次5首进行推送,因此可以获取所述终端标识上传的对本批推荐业务数据(即历史推荐业务数据)的第二业务操作,包括下载、跳过、删除、收藏等操作。所述实时层同样可以采用预设的置信模型,并根据该终端标识下的第二业务操作计算出该终端标识对应的第二标签属性,包括对本批推荐业务数据的第二业务操作所体现的音乐流派、歌手偏好、年代偏好等。
需要说明的是,所述第一业务操作和所述第二业务操作可以为相同的业务操作,采用第一和第二进行区分仅为了说明定时和实时的区别。
推荐数据选取单元12,配置为根据所述定时业务属性和所述实时业务属性在所述目标业务对应的业务数据集合中选取至少一个推荐业务数据;
实现中,所述推荐数据选取单元12可以根据所述定时业务属性和所述实时业务属性在所述目标业务对应的业务数据集合中选取至少一个推荐业务数据。在本发明的其他实施例中,所述业务数据集合包括至少一个业务数据,所述至少一个业务数据中每个业务数据均存在所述每个业务数据对应的第三标签属性,例如:针对音乐应用,音乐库中的每一首音乐数据都对应有音乐流派、歌手、年代等第三标签属性,所述推荐数据选取单元12可以在所述至少一个业务数据中获取第三标签属性满足所述定时业务属性和所述实时业务属性的至少一个目标业务数据,所述推荐数据选取单元12可以在所述业务数据集合中将所述第三标签属性满足定时业务属性和所述实时业务属性的所有业务数据选取出来,所述至少一个目标业务数据表示根据所述定时业务属性和所述实时业务属性在所述业务数据集合中选取的所有业务数据。所述推荐数据选取单元12根据定时业务属性和所述实时业务属性并采用预设评分模型,所述预设评分模型可以为GBDT算法模型, 获取所述至少一个目标业务数据中每个目标业务数据的评分数值,所述推荐数据选取单元12按照每个目标业务数据的评分数值的大小,在所述至少一个目标业务数据中获取至少一个推荐业务数据。
在本发明的其他实施例中,请一并参见图5,为本发明实施例提供了推荐数据选取单元的结构示意图。如图5所示,所述推荐数据选取单元12可以包括:
目标数据获取子单元121,配置为在所述至少一个业务数据中获取第三标签属性满足所述定时业务属性和所述实时业务属性的至少一个目标业务数据;
实现中,所述业务数据集合包括至少一个业务数据,所述至少一个业务数据中每个业务数据均存在所述每个业务数据对应的第三标签属性,例如:针对音乐应用,音乐库中的每一首音乐数据都对应有音乐流派、歌手、年代等第三标签属性,所述目标数据获取子单元121可以在所述至少一个业务数据中获取第三标签属性满足所述定时业务属性和所述实时业务属性的至少一个目标业务数据,所述目标数据获取子单元121可以在所述业务数据集合中将第三标签属性满足所述定时业务属性和所述实时业务属性的所有业务数据选取出来,所述至少一个目标业务数据表示根据所述定时业务属性和所述实时业务属性在所述业务数据集合中选取的所有业务数据。
在本发明的其他实施例中,所述目标数据获取子单元121可以在所述至少一个业务数据中获取与所述定时业务属性匹配的第三标签属性对应的第一候选业务数据,例如:针对音乐应用,在音乐数据集合中存在A音乐数据,其第三标签属性可以为XX流派、YY歌手、ZZ年代,存在B音乐数据,其第三标签属性可以为OO流派、PP歌手、QQ年代,存在C音乐数据,其第三标签属性可以为DD流派、EE歌手、FF年代,所述定时业务属性为XX流派和QQ年代,则所述目标数据获取子单元121可以将A音 乐数据和B音乐数据作为第一候选业务数据。所述目标数据获取子单元121在所述至少一个业务数据中获取与所述实时业务属性匹配的第三标签属性对应的第二候选业务数据,依据上述举例,若实时层检测到所述终端标识执行的第二业务操作包括连续对EE歌手的音乐数据进行收藏,则所述目标数据获取子单元121可以将C音乐数据作为第二候选业务数据。所述目标数据获取子单元121将所述第一候选业务数据和所述第二候选业务数据确定为至少一个目标业务数据。
在本发明的其他实施例中,所述目标数据获取子单元121还可以在所述至少一个业务数据中获取与所述定时业务属性匹配的第三标签属性对应的第三候选业务数据,例如:针对音乐应用,在音乐数据集合中存在A音乐数据,其第三标签属性可以为XX流派、YY歌手、ZZ年代,存在B音乐数据,其第三标签属性可以为OO流派、PP歌手、QQ年代,存在C音乐数据,其第三标签属性可以为DD流派、EE歌手、FF年代,所述定时业务属性为XX流派、PP歌手和FF年代,则所述目标数据获取子单元121可以将A音乐数据、B音乐数据和C音乐数据作为第三候选业务数据。所述目标数据获取子单元121在第三候选业务数据中删除与所述实时业务属性匹配的第三标签属性对应的第四候选业务数据,依据上述举例,若实时层检测到所述终端标识执行的第二业务操作包括连续对PP歌手的音乐数据进行删除,则所述目标数据获取子单元121可以将B音乐数据作为第四候选业务数据,所述目标数据获取子单元121根据删除了所述第四候选业务数据的第三候选业务数据生成至少一个目标业务数据。
数值获取子单元122,配置为根据定时业务属性和所述实时业务属性并采用预设评分模型,获取所述至少一个目标业务数据中每个目标业务数据的评分数值;
实现中,所述数值获取子单元122根据定时业务属性和所述实时业务 属性并采用预设评分模型,所述预设评分模型可以为GBDT算法模型,获取所述至少一个目标业务数据中每个目标业务数据的评分数值,进一步的,所述数值获取子单元122可以将所述定时业务属性和所述实时业务属性作为所述预设评分模型的权重参数,计算所述至少一个目标业务数据中每个目标业务数据的评分数值。
推荐数据获取子单元123,配置为按照每个目标业务数据的评分数值的大小,在所述至少一个目标业务数据中获取至少一个推荐业务数据;
实现中,所述推荐数据获取子单元123按照每个目标业务数据的评分数值的大小,在所述至少一个目标业务数据中获取至少一个推荐业务数据。在本发明的其他实施例中,所述推荐数据获取子单元123可以按照所述每个目标业务数据的评分数值的大小,对所述至少一个目标业务数据进行排序,并在排序后的至少一个目标业务数据中选取预设数量的且评分数值最高的目标业务数据作为至少一个推荐业务数据。
进一步的,所述推荐数据获取子单元123还可以基于预设的筛选规则对选取的预设数量的且评分数值最高的目标业务数据进行再筛选,得到所述至少一个推荐业务数据,所述筛选规则可以为开发人员基于所述目标业务的业务特性设定的筛选规则,例如:针对音乐应用,所述筛选规则可以规定选取的预设数量的且评分数值最高的音乐数据中不能存在相同歌手的标签属性等。
数据发送单元13,配置为将所述至少一个推荐业务数据发送至所述用户终端;
实现中,所述数据发送单元13可以将所述至少一个推荐业务数据发送至所述终端标识对应的所述用户终端。在本发明的其他实施例中,针对上述相对静态的应用场景,所述数据发送单元13可以在接收到所述用户终端发送的携带有所述终端标识的业务数据推荐请求时,查找所述终端标识对 应的所述至少一个推荐业务数据,并将所述至少一个推荐业务数据发送至所述用户终端。
而针对上述动态的应用场景,请一并参见图6,为本发明实施例提供了数据发送单元的结构示意图。如图6所示,所述数据发送单元13可以包括:
数据排序子单元131,配置为按照所述至少一个推荐业务数据中每个推荐业务数据的评分数值,对所述至少一个推荐业务数据进行排序处理;
数据集合划分子单元132,配置为根据排序处理后的至少一个推荐业务数据,将所述至少一个推荐业务数据划分为至少一个推荐业务数据集合;
数据集合发送子单元133,配置为当接收用户终端针对目标业务发送的携带有所述终端标识的业务数据推荐请求时,依次将所述至少一个推荐业务数据集合发送至所述用户终端;
实现中,所述数据排序子单元131可以依据所述至少一个推荐业务数据中每个推荐业务数据的评分数值,对所述至少一个推荐业务数据进行排序处理,所述数据集合划分子单元132在排序处理后,对所述至少一个推荐业务数据进行划分处理,划分为至少一个推荐业务数据集合,所述至少一个推荐业务数据集合中每个推荐业务数据集合包含预设数量的推荐业务数据,例如:针对音乐应用,一共选取15首音乐数据作为一批推荐业务数据,并分3次,每次5首进行推送等,当接收用户终端针对目标业务发送的携带有所述终端标识的业务数据推荐请求时,所述数据集合发送子单元133可以依次将所述至少一个推荐业务数据集合发送至所述用户终端。在本发明的其他实施例中,可以每接收一次业务数据推荐请求,发送一个推荐业务数据集合;或者,可以接收一次业务数据推荐请求后,每隔一段时间发送一个推荐业务数据集合,可以理解的是,本次发送的推荐业务数据集合区别于历史发送的推荐业务数据集合。当然,接收业务数据推荐请求的步骤也可以在获取用户终端的终端标识对应的定时业务属性以及所述终端 标识对应的实时业务属性的步骤之前执行,即接收到业务数据推荐请求,才执行推荐业务数据的选取过程。以上接收业务数据推荐请求的执行顺序可以根据不同的目标业务的业务需求进行决定。
在本发明的其他实施例中,所述数据推荐设备1可以将所述业务数据推荐请求添加至采用分布式内存队列***所构建的请求队列中等待处理,可以理解的是,采用分布式内存队列***构建的请求队列具有轻量级、高性能、易使用、多队列、持久化、分布式容错以及超时控制等特性,当然,也可以采用其它的内存队列***进行请求队列的构建,例如:RabbitMQ、Kafka、Memcacheq或Fqueue等。
在本发明实施例中,通过获取离线层的定时业务属性以及实时层的实时业务属性,在业务数据集合中实时选取推荐业务数据,并推送至用户终端。实现了基于离线层和实时层提供的业务属性,实时生成推荐业务数据的过程,提高了推荐的业务数据的更新效率,提升了业务数据推荐的效果,同时分担了离线层的业务数据选取以及实时层的业务数据推荐的工作,进而提升了业务数据推荐的工作效率;通过基于定时业务属性以及实时业务属性实现推荐业务数据的选取,满足了不同的静态以及动态的应用场景,提升了用户体验;通过采用分布式内存队列***构建请求队列,可以使得请求队列具备轻量级、高性能、易使用、多队列、持久化、分布式容错以及超时控制等特性。
本发明实施例中数据推荐设备所包括的各单元,以及各单元所包括的各子单元都可以通过数据推荐设备中处理器来实现;其中处理器所实现的功能当然还可以通过逻辑电路来实现,在实施的过程中,处理器可以为中央处理器(CPU)、微处理器(MPU)、数字信号处理器(DSP)或现场可编程门阵列(FPGA)等。
请参见图7,为本发明实施例提供了另一种数据推荐设备的结构示意 图。如图7所示,所述数据推荐设备1000可以包括:至少一个处理器1001,例如CPU,至少一个网络接口1004,用户接口1003,存储器1005,至少一个通信总线1002。其中,通信总线1002用于实现这些组件之间的连接通信。其中,用户接口1003可以包括显示屏(Display)、键盘(Keyboard),用户接口1003还可以包括标准的有线接口、无线接口。网络接口1004可以包括标准的有线接口、无线接口(如WI-FI接口)。存储器1005可以是高速RAM存储器,也可以是非不稳定的存储器(non-volatile memory),例如至少一个磁盘存储器。存储器1005还可以是至少一个位于远离前述处理器1001的存储装置。如图7所示,作为一种计算机存储介质的存储器1005中可以包括操作***、网络通信模块、用户接口模块以及数据推荐应用程序。
在图7所示的数据推荐设备1000中,用户接口1003配置为为用户提供输入的接口,获取用户输入的数据;网络接口1004配置为与用户终端相连接,获取用户终端的数据;而处理器1001可以配置为调用存储器1005中存储的数据推荐应用程序,并执行以下操作:
获取用户终端的终端标识对应的定时业务属性以及所述终端标识对应的实时业务属性,所述定时业务属性为根据所述终端标识在第一预设时间段内上传的针对目标业务的第一业务操作,所计算得到所述终端标识对应的第一标签属性;所述实时业务属性为根据所述终端标识实时上传的针对所述目标业务的第二业务操作,所计算得到的所述终端标识对应的第二标签属性;
根据所述定时业务属性和所述实时业务属性在所述目标业务对应的业务数据集合中选取至少一个推荐业务数据;
将所述至少一个推荐业务数据发送至所述用户终端。
在一个实施例中,所述处理器1001在执行获取用户终端的终端标识对 应的定时业务属性以及所述终端标识对应的实时业务属性时,执行以下操作:
获取基于记录的所述终端标识在第一预设时间段内上传的针对目标业务的第一业务操作,并采用预设的推荐算法模型计算得到的所述终端标识对应的定时业务属性;
获取基于记录的所述终端标识在第二预设时间段内上传的针对目标业务的第二业务操作,并采用预设的置信模型计算得到的所述终端标识对应的实时业务属性;
其中,所述第一预设时间段大于所述第二预设时间段。
在一个实施例中,所述处理器1001在执行获取用户终端的终端标识对应的定时业务属性以及所述终端标识对应的实时业务属性时,执行以下操作:
获取基于记录的所述终端标识在第一预设时间段内上传的针对目标业务的第一业务操作,并采用预设的推荐算法模型计算得到的所述终端标识对应的定时业务属性;
获取基于记录的所述终端标识对历史推荐业务数据的第二业务操作,并采用预设的置信模型计算得到的所述终端标识对应的实时业务属性。
在一个实施例中,所述业务数据集合包括至少一个业务数据,所述至少一个业务数据中每个业务数据均存在所述每个业务数据对应的第三标签属性;
所述处理器1001在执行根据所述定时业务属性和所述实时业务属性在所述目标业务对应的业务数据集合中选取至少一个推荐业务数据时,执行以下操作:
在所述至少一个业务数据中获取第三标签属性满足所述定时业务属性和所述实时业务属性的至少一个目标业务数据;
根据定时业务属性和所述实时业务属性并采用预设评分模型,获取所述至少一个目标业务数据中每个目标业务数据的评分数值;
按照每个目标业务数据的评分数值的大小,在所述至少一个目标业务数据中获取至少一个推荐业务数据。
在一个实施例中,所述处理器1001在执行在所述至少一个业务数据中获取第三标签属性满足所述定时业务属性和所述实时业务属性的至少一个目标业务数据时,执行以下操作:
在所述至少一个业务数据中获取与所述定时业务属性匹配的第三标签属性对应的第一候选业务数据;
在所述至少一个业务数据中获取与所述实时业务属性匹配的第三标签属性对应的第二候选业务数据;
将所述第一候选业务数据和所述第二候选业务数据确定为至少一个目标业务数据。
在一个实施例中,所述处理器1001在执行在所述至少一个业务数据中获取第三标签属性满足所述定时业务属性和所述实时业务属性的至少一个目标业务数据时,执行以下操作:
在所述至少一个业务数据中获取与所述定时业务属性匹配的第三标签属性对应的第三候选业务数据;
在第三候选业务数据中删除与所述实时业务属性匹配的第三标签属性对应的第四候选业务数据;
根据删除了所述第四候选业务数据的第三候选业务数据生成至少一个目标业务数据。
在一个实施例中,所述处理器1001在执行将所述至少一个推荐业务数据发送至所述用户终端时,执行以下操作:
按照所述至少一个推荐业务数据中每个推荐业务数据的评分数值,对 所述至少一个推荐业务数据进行排序处理;
根据排序处理后的至少一个推荐业务数据,将所述至少一个推荐业务数据划分为至少一个推荐业务数据集合,所述至少一个推荐业务数据集合中每个推荐业务数据集合包含预设数量的推荐业务数据;
当接收用户终端针对目标业务发送的携带有所述终端标识的业务数据推荐请求时,依次将所述至少一个推荐业务数据集合发送至所述用户终端。
在本发明实施例中,通过获取离线层的定时业务属性以及实时层的实时业务属性,在业务数据集合中实时选取推荐业务数据,并推送至用户终端。实现了基于离线层和实时层提供的业务属性,实时生成推荐业务数据的过程,提高了推荐的业务数据的更新效率,提升了业务数据推荐的效果,同时分担了离线层的业务数据选取以及实时层的业务数据推荐的工作,进而提升了业务数据推荐的工作效率;通过基于定时业务属性以及实时业务属性实现推荐业务数据的选取,满足了不同的静态以及动态的应用场景,提升了用户体验。
需要说明的是,本发明实施例中,如果以软件功能模块的形式实现上述的数据推荐方法,并作为独立的产品销售或使用时,也可以存储在一个计算机可读取存储介质中。基于这样的理解,本发明实施例的技术方案本质上或者说对现有技术做出贡献的部分可以以软件产品的形式体现出来,该计算机软件产品存储在一个存储介质中,包括若干指令用以使得一台计算机设备(可以是个人计算机、服务器、或者网络设备等)执行本发明各个实施例所述方法的全部或部分。而前述的存储介质包括:U盘、移动硬盘、只读存储器(ROM,Read Only Memory)、磁碟或者光盘等各种可以存储程序代码的介质。这样,本发明实施例不限制于任何特定的硬件和软件结合。
相应地,本发明实施例再提供一种计算机存储介质,所述计算机存储 介质中存储有计算机可执行指令,该计算机可执行指令用于执行本发明实施例中数据推荐方法。
本领域普通技术人员可以理解实现上述实施例方法中的全部或部分流程,是可以通过计算机程序来指令相关的硬件来完成,所述的程序可存储于一计算机可读取存储介质中,该程序在执行时,可包括如上述各方法的实施例的流程。其中,所述的存储介质可为磁碟、光盘、只读存储记忆体(Read-Only Memory,ROM)或随机存储记忆体(Random Access Memory,RAM)等。
以上所揭露的仅为本发明较佳实施例而已,当然不能以此来限定本发明之权利范围,因此依本发明权利要求所作的等同变化,仍属本发明所涵盖的范围。发明的保护范围之内。
工业实用性
本发明实施例中:获取用户终端的终端标识对应的定时业务属性以及终端标识对应的实时业务属性,定时业务属性为根据终端标识在第一预设时间段内上传的针对目标业务的第一业务操作,所计算得到终端标识对应的第一标签属性;实时业务属性为根据终端标识实时上传的针对目标业务的第二业务操作,所计算得到的终端标识对应的第二标签属性;根据定时业务属性和实时业务属性在目标业务对应的业务数据集合中选取至少一个推荐业务数据;将至少一个推荐业务数据发送至用户终端。采用本发明实施例提供的技术方案,可以基于离线层和实时层提供的业务属性,实时生成推荐业务数据,提高推荐的业务数据的更新效率,提升业务数据推荐的效果。

Claims (16)

  1. 一种数据推荐方法,包括:
    获取用户终端的终端标识对应的定时业务属性以及所述终端标识对应的实时业务属性,所述定时业务属性为根据所述终端标识在第一预设时间段内上传的针对目标业务的第一业务操作,所计算得到所述终端标识对应的第一标签属性;所述实时业务属性为根据所述终端标识实时上传的针对所述目标业务的第二业务操作,所计算得到的所述终端标识对应的第二标签属性;
    根据所述定时业务属性和所述实时业务属性在所述目标业务对应的业务数据集合中选取至少一个推荐业务数据;
    将所述至少一个推荐业务数据发送至所述用户终端。
  2. 根据权利要求1所述的方法,其中,所述获取用户终端的终端标识对应的定时业务属性以及所述终端标识对应的实时业务属性,包括:
    获取基于记录的所述终端标识在第一预设时间段内上传的针对目标业务的第一业务操作,并采用预设的推荐算法模型计算得到的所述终端标识对应的定时业务属性;
    获取基于记录的所述终端标识在第二预设时间段内上传的针对目标业务的第二业务操作,并采用预设的置信模型计算得到的所述终端标识对应的实时业务属性;
    其中,所述第一预设时间段大于所述第二预设时间段。
  3. 根据权利要求1所述的方法,其中,所述获取用户终端的终端标识对应的定时业务属性以及所述终端标识对应的实时业务属性,包括:
    获取基于记录的所述终端标识在第一预设时间段内上传的针对目标业务的第一业务操作,并采用预设的推荐算法模型计算得到的所述终端标识对应的定时业务属性;
    获取基于记录的所述终端标识对历史推荐业务数据的第二业务操作,并采用预设的置信模型计算得到的所述终端标识对应的实时业务属性。
  4. 根据权利要求1所述的方法,其中,所述业务数据集合包括至少一个业务数据,所述至少一个业务数据中每个业务数据均存在对应的第三标签属性;
    所述根据所述定时业务属性和所述实时业务属性在所述目标业务对应的业务数据集合中选取至少一个推荐业务数据,包括:
    在所述至少一个业务数据中获取第三标签属性满足所述定时业务属性和所述实时业务属性的至少一个目标业务数据;
    根据定时业务属性和所述实时业务属性并采用预设评分模型,获取所述至少一个目标业务数据中每个目标业务数据的评分数值;
    按照每个目标业务数据的评分数值的大小,在所述至少一个目标业务数据中获取至少一个推荐业务数据。
  5. 根据权利要求4所述的方法,其中,所述在所述至少一个业务数据中获取第三标签属性满足所述定时业务属性和所述实时业务属性的至少一个目标业务数据,包括:
    在所述至少一个业务数据中获取与所述定时业务属性匹配的第三标签属性对应的第一候选业务数据;
    在所述至少一个业务数据中获取与所述实时业务属性匹配的第三标签属性对应的第二候选业务数据;
    将所述第一候选业务数据和所述第二候选业务数据确定为至少一个目标业务数据。
  6. 根据权利要求4所述的方法,其中,所述在所述至少一个业务数据中获取第三标签属性满足所述定时业务属性和所述实时业务属性的至 少一个目标业务数据,包括:
    在所述至少一个业务数据中获取与所述定时业务属性匹配的第三标签属性对应的第三候选业务数据;
    在第三候选业务数据中删除与所述实时业务属性匹配的第三标签属性对应的第四候选业务数据;
    根据删除了所述第四候选业务数据的第三候选业务数据生成至少一个目标业务数据。
  7. 根据权利要求4所述的方法,其中,所述将所述至少一个推荐业务数据发送至所述用户终端,包括:
    按照所述至少一个推荐业务数据中每个推荐业务数据的评分数值,对所述至少一个推荐业务数据进行排序处理;
    根据排序处理后的至少一个推荐业务数据,将所述至少一个推荐业务数据划分为至少一个推荐业务数据集合,所述至少一个推荐业务数据集合中每个推荐业务数据集合包含预设数量的推荐业务数据;
    当接收用户终端针对目标业务发送的携带有所述终端标识的业务数据推荐请求时,依次将所述至少一个推荐业务数据集合发送至所述用户终端。
  8. 一种数据推荐设备,包括:
    属性获取单元,配置为获取用户终端的终端标识对应的定时业务属性以及所述终端标识对应的实时业务属性,所述定时业务属性为根据所述终端标识在第一预设时间段内上传的针对目标业务的第一业务操作,所计算得到所述终端标识对应的第一标签属性;所述实时业务属性为根据所述终端标识实时上传的针对所述目标业务的第二业务操作,所计算得到的所述终端标识对应的第二标签属性;
    推荐数据选取单元,配置为根据所述定时业务属性和所述实时业务 属性在所述目标业务对应的业务数据集合中选取至少一个推荐业务数据;
    数据发送单元,配置为将所述至少一个推荐业务数据发送至所述用户终端。
  9. 根据权利要求8所述的设备,其中,所述属性获取单元配置为:
    获取基于记录的所述终端标识在第一预设时间段内上传的针对目标业务的第一业务操作,并采用预设的推荐算法模型计算得到的所述终端标识对应的定时业务属性;
    获取基于记录的所述终端标识在第二预设时间段内上传的针对目标业务的第二业务操作,并采用预设的置信模型计算得到的所述终端标识对应的实时业务属性;
    其中,所述第一预设时间段大于所述第二预设时间段。
  10. 根据权利要求8所述的设备,其中,所述属性获取单元配置为:
    获取基于记录的所述终端标识在第一预设时间段内上传的针对目标业务的第一业务操作,并采用预设的推荐算法模型计算得到的所述终端标识对应的定时业务属性;
    获取基于记录的所述终端标识对历史推荐业务数据的第二业务操作,并采用预设的置信模型计算得到的所述终端标识对应的实时业务属性。
  11. 根据权利要求8所述的设备,其中,所述业务数据集合包括至少一个业务数据,所述至少一个业务数据中每个业务数据均存在所述每个业务数据对应的第三标签属性;
    所述推荐数据选取单元包括:
    目标数据获取子单元,配置为在所述至少一个业务数据中获取第三标签属性满足所述定时业务属性和所述实时业务属性的至少一个目标业 务数据;
    数值获取子单元,配置为根据定时业务属性和所述实时业务属性并采用预设评分模型,获取所述至少一个目标业务数据中每个目标业务数据的评分数值;
    推荐数据获取子单元,配置为按照每个目标业务数据的评分数值的大小,在所述至少一个目标业务数据中获取至少一个推荐业务数据。
  12. 根据权利要求11所述的设备,其中,所述目标数据获取子单元配置为:
    在所述至少一个业务数据中获取与所述定时业务属性匹配的第三标签属性对应的第一候选业务数据;
    在所述至少一个业务数据中获取与所述实时业务属性匹配的第三标签属性对应的第二候选业务数据;
    将所述第一候选业务数据和所述第二候选业务数据确定为至少一个目标业务数据。
  13. 根据权利要求11所述的设备,其中,所述目标数据获取子单元配置为:
    在所述至少一个业务数据中获取与所述定时业务属性匹配的第三标签属性对应的第三候选业务数据;
    在第三候选业务数据中删除与所述实时业务属性匹配的第三标签属性对应的第四候选业务数据;
    根据删除了所述第四候选业务数据的第三候选业务数据生成至少一个目标业务数据。
  14. 根据权利要求11所述的设备,其中,所述数据发送单元包括:
    数据排序子单元,配置为按照所述至少一个推荐业务数据中每个推荐业务数据的评分数值,对所述至少一个推荐业务数据进行排序处理;
    数据集合划分子单元,配置为根据排序处理后的至少一个推荐业务数据,将所述至少一个推荐业务数据划分为至少一个推荐业务数据集合,所述至少一个推荐业务数据集合中每个推荐业务数据集合包含预设数量的推荐业务数据;
    数据集合发送子单元,配置为当接收用户终端针对目标业务发送的携带有所述终端标识的业务数据推荐请求时,依次将所述至少一个推荐业务数据集合发送至所述用户终端。
  15. 一种计算机存储介质,所述计算机存储介质中存储有计算机可执行指令,该计算机可执行指令用于执行权利要求1至7任一项所述的数据推荐方法。
  16. 一种数据推荐设备,包括:
    存储介质,配置为存储可执行指令;
    处理器,配置为执行存储的可执行指令,所述可执行指令用于执行下面的步骤:
    获取用户终端的终端标识对应的定时业务属性以及所述终端标识对应的实时业务属性,所述定时业务属性为根据所述终端标识在第一预设时间段内上传的针对目标业务的第一业务操作,所计算得到所述终端标识对应的第一标签属性;所述实时业务属性为根据所述终端标识实时上传的针对所述目标业务的第二业务操作,所计算得到的所述终端标识对应的第二标签属性;
    根据所述定时业务属性和所述实时业务属性在所述目标业务对应的业务数据集合中选取至少一个推荐业务数据;
    将所述至少一个推荐业务数据发送至所述用户终端。
PCT/CN2017/083524 2016-05-12 2017-05-08 一种数据推荐方法及其设备、存储介质 WO2017193897A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US16/049,352 US10706363B2 (en) 2016-05-12 2018-07-30 Data recommendation method and device, and storage medium

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201610316826.8 2016-05-12
CN201610316826.8A CN105959374B (zh) 2016-05-12 2016-05-12 一种数据推荐方法及其设备

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US16/049,352 Continuation-In-Part US10706363B2 (en) 2016-05-12 2018-07-30 Data recommendation method and device, and storage medium

Publications (1)

Publication Number Publication Date
WO2017193897A1 true WO2017193897A1 (zh) 2017-11-16

Family

ID=56912619

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2017/083524 WO2017193897A1 (zh) 2016-05-12 2017-05-08 一种数据推荐方法及其设备、存储介质

Country Status (3)

Country Link
US (1) US10706363B2 (zh)
CN (1) CN105959374B (zh)
WO (1) WO2017193897A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110852830A (zh) * 2019-09-30 2020-02-28 口碑(上海)信息技术有限公司 一种目标用户信息的获取方法与一种服务订阅***
CN111913996A (zh) * 2020-07-14 2020-11-10 中国联合网络通信集团有限公司 数据处理方法、装置、设备及存储介质

Families Citing this family (26)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105959374B (zh) 2016-05-12 2019-05-03 腾讯科技(深圳)有限公司 一种数据推荐方法及其设备
CN108304429B (zh) * 2017-05-16 2021-12-14 腾讯科技(深圳)有限公司 信息推荐方法、装置和计算机设备
CN108197327B (zh) * 2018-02-07 2020-07-31 腾讯音乐娱乐(深圳)有限公司 歌曲推荐方法、装置及存储介质
CN110555135B (zh) * 2018-03-27 2023-04-07 阿里巴巴(中国)有限公司 内容推荐方法、内容推荐装置和电子设备
CN110555157B (zh) * 2018-03-27 2023-04-07 阿里巴巴(中国)有限公司 内容推荐方法、内容推荐装置和电子设备
CN110555131B (zh) * 2018-03-27 2023-04-07 阿里巴巴(中国)有限公司 内容推荐方法、内容推荐装置和电子设备
CN108696457B (zh) * 2018-04-11 2021-04-23 北京三快在线科技有限公司 一种业务流量分发方法及装置、电子设备及存储介质
US10862811B1 (en) * 2018-06-08 2020-12-08 West Corporation Message brokering for asynchronous status updates
CN109067838B (zh) * 2018-06-29 2021-10-19 聚好看科技股份有限公司 一种数据的推送方法和装置
US11030493B2 (en) * 2018-09-20 2021-06-08 International Business Machines Corporation Estimating sequential blood-sugar levels using images of meals
CN109522337A (zh) * 2018-10-19 2019-03-26 中国平安人寿保险股份有限公司 电子装置、数据整理推送方法和计算机可读存储介质
CN111724187A (zh) * 2019-03-21 2020-09-29 上海晶赞融宣科技有限公司 Dmp受众数据实时处理方法及装置、计算机可读存储介质
CN110162518B (zh) * 2019-04-16 2023-10-31 平安科技(深圳)有限公司 数据分组方法、装置、电子设备及存储介质
CN111861605A (zh) * 2019-04-28 2020-10-30 阿里巴巴集团控股有限公司 业务对象推荐方法
CN111914159B (zh) * 2019-05-10 2024-03-12 招商证券股份有限公司 一种信息推荐方法及终端
CN110197331A (zh) * 2019-05-24 2019-09-03 深圳前海微众银行股份有限公司 业务数据处理方法、装置、设备及计算机可读存储介质
CN110309360B (zh) * 2019-06-13 2021-09-28 山东大学 短视频标签标注方法及***
CN112182358B (zh) * 2019-07-05 2024-04-30 百度在线网络技术(北京)有限公司 一种多媒体推送计划的创建方法和创建***
CN111026969B (zh) * 2019-12-18 2023-10-03 腾讯科技(深圳)有限公司 一种内容推荐方法、装置及存储介质和服务器
CN112306953A (zh) * 2020-01-14 2021-02-02 深圳新阳蓝光能源科技股份有限公司 一种智慧城市数据归档方法、装置和电子设备
CN111737344B (zh) * 2020-05-20 2023-08-25 时时同云科技(成都)有限责任公司 一种针对标签式称重设备的数据同步方法及装置
CN111723287B (zh) * 2020-06-03 2021-03-05 开普云信息科技股份有限公司 一种基于大规模机器学习的内容和服务推荐方法及其***
CN113010536B (zh) * 2021-04-21 2024-02-06 中国平安人寿保险股份有限公司 基于流式数据处理的用户标签获取方法及装置
CN113139833B (zh) * 2021-04-29 2022-10-18 杭州弧途科技有限公司 一种基于用户活跃时区预测和流量分发优化的推荐方法
CN113536115A (zh) * 2021-06-21 2021-10-22 浙江吉利控股集团有限公司 一种网约车信息推送方法、装置及平台
US11663216B2 (en) * 2021-07-28 2023-05-30 Bank Of America Corporation Delta database data provisioning

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120066393A1 (en) * 2010-09-15 2012-03-15 Core Mobile Networks, Inc. System and method for real time delivery of context based content from the cloud to mobile devices
CN103118111A (zh) * 2013-01-31 2013-05-22 北京百分点信息科技有限公司 一种基于多个数据交互中心的数据进行信息推送的方法
CN104090894A (zh) * 2013-12-18 2014-10-08 深圳市腾讯计算机***有限公司 在线并行计算推荐信息的方法、装置及服务器
CN104424210A (zh) * 2013-08-22 2015-03-18 腾讯科技(深圳)有限公司 一种信息推荐方法、***及服务器
CN105959374A (zh) * 2016-05-12 2016-09-21 腾讯科技(深圳)有限公司 一种数据推荐方法及其设备

Family Cites Families (17)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20010028662A1 (en) * 2000-01-18 2001-10-11 Hunt Paul M. Method and system of real-time optimization and implementation of content and advertising programming decisions for broadcasts and narrowcasts
EP1156424A2 (en) * 2000-05-17 2001-11-21 Matsushita Electric Industrial Co., Ltd. Information recommendation apparatus and information recommendation system
US20040193938A1 (en) * 2003-03-28 2004-09-30 Shah Rasiklal Punjalal Complex system serviceability method and apparatus
US20120173370A1 (en) * 2005-09-14 2012-07-05 Adam Soroca System for retrieving mobile communication facility user data from a plurality of providers
US20090163183A1 (en) * 2007-10-04 2009-06-25 O'donoghue Hugh Recommendation generation systems, apparatus and methods
US20110125585A1 (en) * 2009-11-20 2011-05-26 Rovi Technologies Corporation Content recommendation for a content system
US20120310770A1 (en) * 2011-06-01 2012-12-06 Universidade Da Coruna (Otri) Recommendation method and system
US9390181B1 (en) * 2011-06-09 2016-07-12 Amazon Technologies, Inc. Personalized landing pages
WO2014004735A1 (en) * 2012-06-26 2014-01-03 Medio Systems, Inc. Recommendations system
CN103905507A (zh) * 2012-12-28 2014-07-02 ***通信集团公司 一种业务信息推荐***及业务信息推荐方法
US20150006286A1 (en) * 2013-06-28 2015-01-01 Linkedin Corporation Targeting users based on categorical content interactions
US20150006294A1 (en) * 2013-06-28 2015-01-01 Linkedln Corporation Targeting rules based on previous recommendations
US11809501B2 (en) * 2014-08-28 2023-11-07 Ebay Inc. Systems, apparatuses, and methods for providing a ranking based recommendation
US11675833B2 (en) * 2015-12-30 2023-06-13 Yahoo Assets Llc Method and system for recommending content
US20170289202A1 (en) * 2016-03-31 2017-10-05 Microsoft Technology Licensing, Llc Interactive online music experience
US9779443B1 (en) * 2016-05-27 2017-10-03 Blueshift Labs, Inc. Event-based personalized merchandising schemes and applications in messaging
US10943263B2 (en) * 2016-12-21 2021-03-09 Meru Cab Company Private Limited System, method, and computer-readable medium for providing optimal dynamic content of a service to a user

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120066393A1 (en) * 2010-09-15 2012-03-15 Core Mobile Networks, Inc. System and method for real time delivery of context based content from the cloud to mobile devices
CN103118111A (zh) * 2013-01-31 2013-05-22 北京百分点信息科技有限公司 一种基于多个数据交互中心的数据进行信息推送的方法
CN104424210A (zh) * 2013-08-22 2015-03-18 腾讯科技(深圳)有限公司 一种信息推荐方法、***及服务器
CN104090894A (zh) * 2013-12-18 2014-10-08 深圳市腾讯计算机***有限公司 在线并行计算推荐信息的方法、装置及服务器
CN105959374A (zh) * 2016-05-12 2016-09-21 腾讯科技(深圳)有限公司 一种数据推荐方法及其设备

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110852830A (zh) * 2019-09-30 2020-02-28 口碑(上海)信息技术有限公司 一种目标用户信息的获取方法与一种服务订阅***
CN111913996A (zh) * 2020-07-14 2020-11-10 中国联合网络通信集团有限公司 数据处理方法、装置、设备及存储介质
CN111913996B (zh) * 2020-07-14 2023-07-18 中国联合网络通信集团有限公司 数据处理方法、装置、设备及存储介质

Also Published As

Publication number Publication date
CN105959374A (zh) 2016-09-21
CN105959374B (zh) 2019-05-03
US10706363B2 (en) 2020-07-07
US20180365577A1 (en) 2018-12-20

Similar Documents

Publication Publication Date Title
WO2017193897A1 (zh) 一种数据推荐方法及其设备、存储介质
CN111680221B (zh) 信息推荐方法、装置、设备及计算机可读存储介质
US11971925B2 (en) Predicting topics of potential relevance based on retrieved/created digital media files
WO2017071251A1 (zh) 信息推送方法和装置
WO2019114423A1 (zh) 对模型预测值进行融合的方法、装置和设备
CN110413867B (zh) 用于内容推荐的方法及***
US20120246302A1 (en) System and methodology for creating and using contextual user profiles
CN112528147B (zh) 内容推荐方法和装置、训练方法、计算设备和存储介质
CN112559896B (zh) 信息推荐方法、装置、设备及计算机可读存储介质
CN109377401B (zh) 一种数据处理方法、装置、***、服务器及存储介质
CN112100221A (zh) 一种资讯推荐方法、装置、推荐服务器及存储介质
JP2017059057A (ja) 推定装置、推定方法及び推定プログラム
CN113220974B (zh) 点击率预测模型训练、搜索召回方法、装置、设备及介质
CN113535939A (zh) 文本处理方法和装置、电子设备以及计算机可读存储介质
CN113076450B (zh) 一种目标推荐列表的确定方法和装置
JP6129815B2 (ja) 情報処理装置、方法及びプログラム
JP6050800B2 (ja) 情報処理装置、方法及びプログラム
Bailer et al. Learning selection of user generated event videos
CN111241386A (zh) 一种信息处理方法及装置、存储介质
CN114155049B (zh) 一种确定目标对象的方法和装置
KR102674099B1 (ko) 영상 콘텐츠의 분류 및 시청 관련 데이터의 분석 결과를 기반으로 한 콘텐츠 기획 및 a/b 테스트 자동화 방법, 장치 및 시스템
EP4270291A1 (en) Method and computer device to recommend related product technical field
KR20230153883A (ko) 연관 상품 추천을 위한 방법, 컴퓨터 장치, 및 컴퓨터 프로그램
CN114741502A (zh) 资源确定的方法和装置、电子设备和存储介质
KR20240040322A (ko) 개인화된 예상 평점을 기초로 콘텐츠를 추천하는 방법, 컴퓨터 장치, 및 컴퓨터 프로그램

Legal Events

Date Code Title Description
NENP Non-entry into the national phase

Ref country code: DE

121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 17795515

Country of ref document: EP

Kind code of ref document: A1

122 Ep: pct application non-entry in european phase

Ref document number: 17795515

Country of ref document: EP

Kind code of ref document: A1