CN116975455B - User interest recognition method and device - Google Patents
User interest recognition method and device Download PDFInfo
- Publication number
- CN116975455B CN116975455B CN202311234324.7A CN202311234324A CN116975455B CN 116975455 B CN116975455 B CN 116975455B CN 202311234324 A CN202311234324 A CN 202311234324A CN 116975455 B CN116975455 B CN 116975455B
- Authority
- CN
- China
- Prior art keywords
- page
- push
- interest
- target
- user
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 53
- 238000012545 processing Methods 0.000 claims abstract description 79
- 230000002452 interceptive effect Effects 0.000 claims abstract description 48
- 238000005065 mining Methods 0.000 claims abstract description 28
- 230000006399 behavior Effects 0.000 claims description 196
- 230000003993 interaction Effects 0.000 claims description 144
- 238000012512 characterization method Methods 0.000 claims description 16
- 230000008569 process Effects 0.000 claims description 16
- 125000004122 cyclic group Chemical group 0.000 claims description 13
- 230000002776 aggregation Effects 0.000 claims description 10
- 238000004220 aggregation Methods 0.000 claims description 10
- 230000006872 improvement Effects 0.000 claims description 7
- 230000004931 aggregating effect Effects 0.000 claims description 6
- 238000012549 training Methods 0.000 claims description 6
- 238000004590 computer program Methods 0.000 claims description 4
- 238000011002 quantification Methods 0.000 claims description 4
- 238000003860 storage Methods 0.000 claims description 4
- 238000009412 basement excavation Methods 0.000 claims description 3
- 238000012216 screening Methods 0.000 claims description 3
- 238000004513 sizing Methods 0.000 claims description 3
- 238000005259 measurement Methods 0.000 claims description 2
- 238000004458 analytical method Methods 0.000 abstract description 4
- 238000013461 design Methods 0.000 description 5
- 238000004891 communication Methods 0.000 description 4
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 238000013473 artificial intelligence Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000007499 fusion processing Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/50—Network services
- H04L67/55—Push-based network services
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Business, Economics & Management (AREA)
- Strategic Management (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Accounting & Taxation (AREA)
- Development Economics (AREA)
- Finance (AREA)
- Databases & Information Systems (AREA)
- General Engineering & Computer Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Marketing (AREA)
- Bioinformatics & Computational Biology (AREA)
- Game Theory and Decision Science (AREA)
- General Business, Economics & Management (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Economics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
According to the user interest identification method and device, in view of the fact that the active debugging examples and the passive debugging examples of the push interest identification network application are priori user page interactive behavior images in the target page service session, the information bearing capacity is rich, and the method and device can comprise all target page attention item information in the target page service session, so that mining analysis of page attention items is facilitated. The push interest recognition networks for completing debugging can recognize target page attention matters from different ideas so as to accurately, comprehensively and rapidly determine the push interest item recognition views of the target page attention matters, and further improve the push interest mining quality of page service users aiming at the page service users to be subjected to push interest recognition processing.
Description
Technical Field
The invention relates to the technical field of artificial intelligence, in particular to a user interest identification method and device.
Background
With the continuous development of the internet, data explosion increases, and the application of data mining technology becomes wider and wider. Among them, user interest mining and application in a big data environment is a very important field.
The user interest mining refers to predicting the interest points, behavior habits and future demands of a user by analyzing information such as user behaviors, interest preferences and historical data of the user, and aims to provide more personalized services, products and information, so that user experience is improved, marketing cost is reduced, and enterprise income is improved.
As one of key technologies of user interest mining, an artificial intelligence algorithm is increasingly widely applied to user interest mining, but in some scenes, how to realize accurate, comprehensive and efficient user interest mining is still a technical problem to be solved at present.
Disclosure of Invention
The invention at least provides a user interest identification method and a device.
The invention provides a user interest identification method which is applied to a user interest identification device, and comprises the following steps:
obtaining a target user page interactive behavior image of a page service user to be subjected to push interest identification processing, and obtaining a push interest identification network set, wherein the push interest identification network set comprises a plurality of push interest identification networks for completing debugging;
mining corresponding interactive behavior state representation knowledge sets from the target user page interactive behavior images through each push interest recognition network, and obtaining confidence coefficients of target page concerns of the page service users to be subjected to push interest recognition processing, which are generated by each push interest recognition network, according to the interactive behavior state representation knowledge sets; the method comprises the steps that positive debugging examples and negative debugging examples of all pushing interest identification network applications are priori user page interactive behavior images in a target page service session link, and the positive debugging examples and the negative debugging examples of different pushing interest identification network applications are different;
And determining that the page service user to be subjected to the push interest identification processing has a push interest item identification view of the target page attention item according to the confidence coefficient generated by each push interest identification network.
In some examples, the respective push interest recognition networks include at least one of:
the first push interest recognition network is used for carrying out push interest recognition through a commonality measurement value between the interactive behavior images of the user pages;
the second push interest recognition network carries out push interest recognition through a plurality of confidence coefficient predicted values corresponding to the user page interaction behavior image; wherein each confidence coefficient predictor is obtained by a target network branch in the second push interest identification network;
and the third push interest recognition network carries out push interest recognition through linkage interaction behavior state characterization knowledge corresponding to the user page interaction behavior image.
In some examples, when the push interest recognition networks are the first push interest recognition networks, mining to obtain corresponding interaction behavior state representation knowledge sets from the target user page interaction behavior images through the push interest recognition networks, and obtaining confidence coefficients of target page concerns of the page service users to be subjected to push interest recognition processing generated by the push interest recognition networks according to the interaction behavior state representation knowledge sets, wherein the confidence coefficients comprise:
According to the first push interest identification network, the following steps are implemented:
screening an auxiliary user page interaction behavior image from the prior user page interaction behavior image, and mining a corresponding auxiliary interaction behavior state representation knowledge set from the auxiliary user page interaction behavior image;
excavating a corresponding first target interaction behavior state representation knowledge set from the target user page interaction behavior image;
determining a push interest commonality score between the target user page interaction behavior image and the auxiliary user page interaction behavior image based on the first target interaction behavior state representation knowledge set and the auxiliary interaction behavior state representation knowledge set;
and generating a first confidence coefficient of the target page attention item existing in the page service user to be subjected to the push interest identification processing based on the push interest commonality score.
In some examples, further comprising:
when the auxiliary user page interaction behavior image is a user page interaction behavior image with a relationship of the target page attention item, a first quantification relationship exists between the push interest commonality score and the first confidence coefficient;
And when the auxiliary user page interaction behavior image is a user page interaction behavior image with a relationship of non-target page attention matters, the push interest commonality score has a second quantitative relationship with the first confidence coefficient.
In some examples, the first push interest recognition network is commissioned based on the following ideas:
obtaining a debugging example binary group set from the prior user page interaction behavior image;
according to the debugging example binary group set, performing repeated cyclic debugging on a first pushing interest recognition network to be debugged, and outputting the first pushing interest recognition network with the completion of the debugging, wherein in the process of each cyclic debugging, the following steps are implemented:
inputting a plurality of selected debugging example tuples into the first pushing interest recognition network to be debugged, and respectively obtaining a debugging example commonality index between two debugging examples contained in each debugging example tuple;
based on the obtained debug example commonality indexes, respectively obtaining page attention item prediction viewpoints of corresponding debug example doublets;
and determining a push interest identification cost variable according to the page attention item authentication viewpoint and the page attention item prediction viewpoint corresponding to each of the plurality of debugging example tuples, and performing network improvement according to the push interest identification cost variable.
In some examples, when the push interest recognition networks are the second push interest recognition networks, mining to obtain corresponding interaction behavior state representation knowledge sets from the target user page interaction behavior images through the push interest recognition networks, and obtaining confidence coefficients of target page concerns of the page service users to be subjected to push interest recognition processing generated by the push interest recognition networks according to the interaction behavior state representation knowledge sets, wherein the confidence coefficients comprise:
respectively mining a second target interaction behavior state representation knowledge set of the target page attention items according to a plurality of target network branches contained in the second push interest identification network;
based on the second target interaction behavior state representation knowledge set corresponding to each of the plurality of target network branches, obtaining a confidence coefficient predicted value of the target user page interaction behavior image generated by the corresponding target network branch, wherein the confidence coefficient predicted value belongs to target page attention matters;
and optimizing a plurality of confidence coefficient predicted values based on the importance degrees of the set page attention items corresponding to the target network branches, and obtaining a second confidence coefficient of the target page attention items of the page service user to be subjected to the push interest identification processing, which is generated by the second push interest identification network.
In some examples, the second push interest recognition network is obtained based on:
obtaining a debugging example set from the prior user page interactive behavior image, wherein the debugging example set comprises a plurality of positive debugging examples and a plurality of negative debugging examples;
predicting network branches according to the set page attention items, and respectively implementing the following steps:
according to the debug example set, obtaining respective page attention item prediction viewpoints of a plurality of debug examples, and determining a local viewpoint prediction cost variable according to respective page attention item authentication viewpoints and page attention item prediction viewpoints corresponding to the plurality of debug examples;
aggregating a plurality of local viewpoint prediction cost variables, and selecting at least two target network branches from the plurality of page attention item prediction network branches based on the target cost variables after aggregation;
integrating a selected minimum of two target network branches into the second push interest identification network.
In some examples, when the push interest recognition networks are the third push interest recognition networks, mining to obtain corresponding interaction behavior state representation knowledge sets from the target user page interaction behavior images through the push interest recognition networks, and obtaining confidence coefficients of target page concerns of the page service users to be subjected to push interest recognition processing generated by the push interest recognition networks according to the interaction behavior state representation knowledge sets, wherein the confidence coefficients comprise:
According to the third pushing interest recognition network, a corresponding third target interaction behavior state representation knowledge set is obtained from the target user page interaction behavior image through excavation;
and performing aggregation operation on each interactive behavior state representation knowledge in the third target interactive behavior state representation knowledge set, and obtaining a third confidence coefficient of target page attention matters of the page service user to be subjected to the push interest identification processing, which is generated by the third push interest identification network, based on the aggregated interactive behavior state representation knowledge.
In some examples, the third push interest recognition network is commissioned based on the following ideas:
obtaining a debugging example set from the prior user page interactive behavior image, wherein the debugging example set comprises a plurality of positive debugging examples and a plurality of negative debugging examples;
according to the debugging example set, performing multiple times of cyclic debugging on a third pushing interest recognition network to be debugged, and outputting the third pushing interest recognition network with the completion of debugging, wherein in each cyclic debugging process, the following steps are implemented:
according to the third pushing interest recognition network to be debugged, obtaining page attention description characteristics of a plurality of debugging examples;
Fusing and sizing the page focus description features of the plurality of debugging examples to obtain fused current page focus description features, and aggregating the current page focus description features with annotation features of the plurality of debugging examples;
based on the combined characteristics of the completion of aggregation, determining a push interest identification cost variable, and carrying out network improvement according to the push interest identification cost variable.
In some examples, when the respective push interest recognition networks include a first push interest recognition network, a second push interest recognition network, and a third push interest recognition network, determining a push interest item recognition perspective of the target page attention item according to respective generated confidence coefficients of the respective push interest recognition networks includes:
when a first confidence coefficient determined by a first push interest identification network is larger than a set first page attention item limit value, determining that a push interest item identification view of the target page attention item exists in the page service user to be subjected to push interest identification processing according to the first confidence coefficient;
when the minimum value of the second confidence coefficient determined by the second push interest recognition network and the minimum value of the third confidence coefficient determined by the third push interest recognition network are both larger than the set second page attention item limit value, determining that the page service user to be subjected to push interest recognition processing has a push interest item recognition view of the target page attention item according to the maximum value of the second confidence coefficient and the third confidence coefficient;
And when the first confidence coefficient is not greater than the first page attention item limit value and the minimum value of the second confidence coefficient and the third confidence coefficient is smaller than or equal to the second page attention item limit value, determining that the page service user to be subjected to push interest identification processing has a push interest item identification view of the target page attention item according to the minimum value of the second confidence coefficient and the third confidence coefficient.
In some examples, when it is determined that the target page attention item exists for the page service user to be subjected to the push interest identification process, the method further includes:
obtaining a target page service session link of the page service user to be subjected to push interest identification processing;
identifying user page interaction behavior images of a plurality of session participants in the target page service session link according to the push interest identification networks;
determining the number of target page attention matters in the target page service session links in a set processing period based on a push interest item identification viewpoint;
and when the number exceeds a set limit value, determining the target page service session link as an active link of the target page attention item.
The invention also provides a user interest recognition device, which comprises a processor and a memory; the processor is in communication with the memory, and the processor is configured to read and execute a computer program from the memory to implement the method described above.
The present invention also provides a computer readable storage medium having stored thereon a computer program which, when run, implements the method described above.
The technical scheme provided by the embodiment of the invention can comprise the following beneficial effects: aiming at the obtained target user page interactive behavior image of the page service user to be subjected to push interest identification processing, identifying by adopting each push interest identification network which completes debugging, determining each confidence coefficient of the target page attention item of the page service user to be subjected to push interest identification processing based on an interactive behavior state representation knowledge set mined by each push interest identification network from the target user page interactive behavior image, and determining the push interest item identification view of the target page attention item of the page service user to be subjected to push interest identification processing based on each confidence coefficient. In view of the fact that the active debugging examples and the passive debugging examples of the push interest recognition network application are priori user page interaction behavior images in the target page service session link, the information bearing capacity is rich, and the push interest recognition network application can comprise all target page attention information in the target page service session link, so that mining analysis of page attention is facilitated. In addition, in view of the prior user page interactive behavior image applied by each push interest recognition network in the debugging period, all interest details of target page attention matters can be included, so that the debugging precision of each push interest recognition network is ensured; compared with the method of calling a single push interest recognition network, the method has the advantages that the active debugging examples and the passive debugging examples of different push interest recognition network applications are different, so that each push interest recognition network which completes debugging can recognize target page attention matters from different ideas, the push interest item recognition views of the target page attention matters can be accurately, comprehensively and quickly determined, and the push interest mining quality of page service users who want to carry out push interest recognition processing is improved.
For a description of the effects of the above user interest recognition means, the computer-readable storage medium, see the description of the above method.
In order to make the above objects, features and advantages of the present invention more comprehensible, preferred embodiments accompanied with figures are described in detail below.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are necessary for the embodiments to be used are briefly described below, the drawings being incorporated in and forming a part of the description, these drawings showing embodiments according to the present invention and together with the description serve to illustrate the technical solutions of the present invention. It is to be understood that the following drawings illustrate only certain embodiments of the invention and are therefore not to be considered limiting of its scope, for the person of ordinary skill in the art may admit to other equally relevant drawings without inventive effort.
Fig. 1 is a block diagram of a user interest recognition apparatus according to an embodiment of the present invention.
Fig. 2 is a flow chart illustrating a method for identifying user interests according to an embodiment of the present invention.
Detailed Description
Reference will now be made in detail to exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, the same numbers in different drawings refer to the same or similar elements, unless otherwise indicated. The implementations described in the following exemplary examples do not represent all implementations consistent with the invention. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the invention.
Fig. 1 is a schematic diagram of a user interest recognition device 10 according to an embodiment of the present invention, including a processor 102, a memory 104, and a bus 106. The memory 104 is used for storing execution instructions, including a memory and an external memory, where the memory may also be understood as an internal memory, and is used for temporarily storing operation data in the processor 102 and data exchanged with the external memory such as a hard disk, where the processor 102 exchanges data with the external memory through the memory, and when the user interest recognition device 10 operates, the processor 102 and the memory 104 communicate with each other through the bus 106, so that the processor 102 executes the user interest recognition method according to the embodiment of the present invention.
Referring to fig. 2, fig. 2 is a schematic flow chart of a user interest recognition method according to an embodiment of the present invention, which is applied to a user interest recognition device, and the method may include steps 1 to 3.
Step 1: the user interest recognition device obtains a target user page interaction behavior image of a page service user to be subjected to push interest recognition processing.
In the embodiment of the invention, the user interest recognition device can acquire the interaction behavior log of the page service user to be subjected to the push interest recognition processing through the page user client, and acquire the user page interaction behavior image of the page service user to be subjected to the push interest recognition processing on the premise of agreeing of the page service user to be subjected to the push interest recognition processing, wherein the user page interaction behavior image can carry out anonymization processing on the individual privacy of the page service user to be subjected to the push interest recognition processing.
The page service may relate to e-commerce, digital office, supply chain finance, intelligent education, etc.
Step 2: the user interest recognition device extracts corresponding interaction behavior state representation knowledge sets from the target user page interaction behavior images through each pushing interest recognition network, and obtains confidence coefficients of target page concerns of page service users to be subjected to pushing interest recognition processing, which are generated by each pushing interest recognition network, according to the interaction behavior state representation knowledge sets.
In the embodiment of the invention, the positive debugging examples and the negative debugging examples of the push interest identification network application are priori user page interactive behavior images in the target page service session link, and the positive debugging examples and the negative debugging examples of different push interest identification network applications are different.
For example, positive debug examples may be positive samples and negative debug examples may be negative samples. The target page service session link may be a page service session within a period of time, or a page service session in a certain session process. The prior user page interaction image may be understood as an authenticated user page interaction image.
Further, a series of operation behaviors of the user in the interactive page, such as clicking behaviors, browsing behaviors, copy-and-paste behaviors, comment behaviors, collection behaviors and the like, are included in the user page interactive behavior image.
When each push interest recognition network in each push interest recognition networks is the first push interest recognition network, the push interest recognition network performs push interest recognition through a commonality metric value between user page interaction behavior images, and a generated process of generating a confidence coefficient of a target page attention item existing in a page service user to be subjected to push interest recognition processing can include the following steps 21-24.
Step 21: the user interest recognition device screens an auxiliary user page interaction behavior image from the priori user page interaction behavior images, and mines a corresponding auxiliary interaction behavior state representation knowledge set from the auxiliary user page interaction behavior image.
Under some exemplary design ideas, the screened auxiliary user page interaction behavior image (reference user page interaction behavior image) can be a user page interaction behavior image with a relationship of target page attention matters, namely, arbitrarily selecting one user page interaction behavior image from the debug sample set as the auxiliary user page interaction behavior image.
In another possible design concept, the screened auxiliary user page interaction behavior image may be a user page interaction behavior image with a relationship of non-target page concerns, in other words, one user page interaction behavior image may be arbitrarily selected from the relevant debug sample set as the auxiliary user page interaction behavior image.
After the auxiliary user page interactive behavior image is screened, in a real-time step 21, an auxiliary interactive behavior state representation knowledge set is obtained by mining from the auxiliary user page interactive behavior image based on the debugged knowledge mining branch of the first push interest recognition network.
Step 22: the user interest recognition device is used for mining and obtaining a corresponding first target interaction behavior state representation knowledge set from the target user page interaction behavior image.
When step 22 is implemented, a first target interaction behavior state representation knowledge set is mined from the target user page interaction behavior image based on the debugged knowledge mining branches of the first push interest recognition network.
Step 23: the user interest recognition device determines a push interest commonality score between the target user page interaction behavior image and the auxiliary user page interaction behavior image based on the first target interaction behavior state representation knowledge set and the auxiliary interaction behavior state representation knowledge set.
Under some exemplary design ideas, the first target interaction behavior state characterization knowledge set and the auxiliary interaction behavior state characterization knowledge set include a series of user interest features corresponding to target page interests, and based on the user interest features, a push interest commonality score between the target user page interaction behavior image and the auxiliary user page interaction behavior image can be determined.
Step 24: and the user interest recognition device generates a first confidence coefficient of the target page attention item existing in the page service user to be subjected to the push interest recognition processing based on the push interest commonality score.
In the embodiment of the invention, the mapping relation between the push interest commonality score and the first confidence coefficient is different based on different screened auxiliary user page interaction behavior images. The push interest commonality score may be understood as a similarity of push interest features.
When the auxiliary user page interaction behavior image is a user page interaction behavior image with a relation to the target page attention item, a first quantification relation (namely positive correlation) exists between the push interest commonality score and the first confidence coefficient, in other words, the greater the push interest commonality score is, the greater the first confidence coefficient of the target page attention item exists for the page service user to be subjected to push interest identification processing is.
For example, when the auxiliary user page interaction image is image1, the determined push interest commonality score Similarity1 is 0.87, the corresponding first confidence coefficient1 is 0.69, and when the auxiliary user page interaction image is image2, the determined push interest commonality score Similarity1 is 0.93, and the corresponding first confidence coefficient1 is 0.81.
When the auxiliary user page interaction behavior image is a user page interaction behavior image with a relationship of non-target page attention items, a second quantitative relationship (namely negative correlation) exists between the push interest commonality score and the first confidence coefficient, in other words, the greater the push interest commonality score is, the lower the first confidence coefficient of the target page attention items exists for the page service user to be subjected to push interest identification processing is.
For another example, when the auxiliary user page interaction image is picture1, the determined push interest commonality score Similarity1 is 0.85, the corresponding first confidence coefficient1 is 0.21, and when the auxiliary user page interaction image is picture2, the determined push interest commonality score Similarity1 is 0.90, and the corresponding first confidence coefficient1 is 0.14.
When each push interest recognition network in each push interest recognition network is a second push interest recognition network, the push interest recognition network carries out push interest recognition through a plurality of confidence coefficient predicted values corresponding to the user page interaction behavior image, each confidence coefficient predicted value is obtained through one target network branch in the second push interest recognition network, and the generated thought of the confidence coefficient of the target page attention item existing in the page service user to be subjected to push interest recognition processing comprises the steps 2 a-2 c.
Step 2a: the user interest recognition device respectively excavates a second target interaction behavior state representation knowledge set of target page attention matters based on a plurality of target network branches contained in the second push interest recognition network.
In the embodiment of the invention, for each target network branch, a second target interaction behavior state representation knowledge set is mined from the target user page interaction behavior image, wherein the second target interaction behavior state representation knowledge set comprises behavior habit characteristics, comment viewpoint characteristics and the like. The second target interaction behavior state representation knowledge set and the first target interaction behavior state representation knowledge set may be consistent or inconsistent.
Step 2b: the user interest recognition device obtains confidence coefficient predicted values of target user page interaction behavior images generated by corresponding target network branches belonging to target page attention matters based on second target interaction behavior state characterization knowledge sets corresponding to the target network branches.
For each target network branch, determining a confidence coefficient predicted value of the target user page interaction behavior image belonging to the target page attention item based on the second target interaction behavior state characterization knowledge set mined by the target network branch. Where a network branch may be understood as a sub-model or sub-network.
Step 2c: the user interest recognition device optimizes (for example, may be a weighting process) the plurality of confidence coefficient predicted values based on the importance degrees of the set page attention items corresponding to the plurality of target network branches, so as to obtain a second confidence coefficient of the target page attention items of the page service user to be subjected to the push interest recognition process, which is generated by the second push interest recognition network.
For example, the confidence coefficient predicted value of the target user page interaction behavior image determined by the target Network Branch1 and belonging to the target page attention item is P1, the confidence coefficient predicted value of the target user page interaction behavior image determined by the target Network Branch2 and belonging to the target page attention item is P2, the confidence coefficient predicted value of the target user page interaction behavior image determined by the target Network Branch3 and belonging to the target page attention item is P3, the importance degrees of the set page attention items corresponding to the target Network branches Network Branch1, network Branch2 and Network Branch3 are Q1, Q2 and Q3 respectively, and the second confidence coefficient cofacient 2 of the target page attention item for the page service user to be subjected to the push interest identification processing generated by the second push interest identification Network is: coeffipresent 2=q1×p1+q2×p2+q3×p3.
When each push interest recognition network in each push interest recognition network is a third push interest recognition network, the push interest recognition network carries out push interest recognition through linkage interaction behavior state characterization knowledge corresponding to a user page interaction behavior image, and the generated thought that the page service user to be subjected to push interest recognition processing has a confidence coefficient of target page attention matters comprises a step 201 and a step 202.
Step 201: the user interest recognition device is based on a third push interest recognition network, and a corresponding third target interaction behavior state representation knowledge set is obtained from the target user page interaction behavior image through excavation.
The third target interaction behavior state characterization knowledge set comprises behavior habit features, comment viewpoint features and the like. The third target interaction behavior state representation knowledge set and the second target interaction behavior state representation knowledge set, and the first target interaction behavior state representation knowledge set may be consistent or inconsistent.
Step 202: the user interest recognition device performs aggregation operation (such as fusion processing) on each interactive behavior state characterization knowledge in the third target interactive behavior state characterization knowledge set, and obtains a third confidence coefficient of target page attention matters of the page service user to be subjected to push interest recognition processing, which is generated by the third push interest recognition network, based on the aggregated interactive behavior state characterization knowledge.
In the embodiment of the invention, the aggregated interactive behavior state characterization knowledge can abundantly characterize the interactive behavior state characteristics of the page service user to be subjected to the push interest identification processing, so that the third confidence coefficient coeffient 3 of the target page attention item of the page service user to be subjected to the push interest identification processing generated by the third push interest identification network is conveniently improved.
The processing of the network identification based on the three push interests does not have a sequence limitation.
Step 3: the user interest recognition device determines the push interest item recognition view of the target page attention item existing in the page service user to be subjected to push interest recognition processing based on the confidence coefficient generated by each push interest recognition network.
The push interest item identification viewpoint may include "no target page attention item exists for the page service user to be subjected to push interest identification processing" and "target page attention item exists for the page service user to be subjected to push interest identification processing".
Under some exemplary design ideas, when step 3 is implemented, based on a confidence coefficient1 generated by the first push interest recognition network, a confidence coefficient2 generated by the second push interest recognition network, and a confidence coefficient3 generated by the third push interest recognition network, it is determined that a page service user to be subjected to push interest recognition processing has a push interest item recognition viewpoint of a target page attention item. For example, when the first confidence coefficient determined by the first push interest recognition network is greater than the set first page attention item limit value, determining that the page service user to be subjected to push interest recognition processing has a push interest item recognition viewpoint of the target page attention item based on the first confidence coefficient; when the minimum value of the second confidence coefficient determined by the second push interest recognition network and the minimum value of the third confidence coefficient determined by the third push interest recognition network are both larger than the set second page attention item limit value, determining that a page service user to be subjected to push interest recognition processing has a push interest item recognition view of a target page attention item based on the maximum value of the second confidence coefficient and the third confidence coefficient; when the first confidence coefficient is not greater than the first page attention item limit value and the minimum value of the second confidence coefficient and the third confidence coefficient is smaller than or equal to the second page attention item limit value, determining that the page service user to be subjected to push interest identification processing has a push interest item identification view of the target page attention item based on the minimum value of the second confidence coefficient and the third confidence coefficient.
In the embodiment of the invention, aiming at the obtained target user page interactive behavior image of the page service user to be subjected to push interest identification processing, each push interest identification network which completes debugging is adopted for identification, each confidence coefficient of target page attention items of the page service user to be subjected to push interest identification processing is determined based on an interactive behavior state representation knowledge set mined from the target user page interactive behavior image by each push interest identification network, and the push interest item identification view of the target page attention items of the page service user to be subjected to push interest identification processing is determined based on each confidence coefficient. In view of the fact that each pushing interest identification network application positive debugging example and each pushing interest identification network application negative debugging example are priori user page interaction behavior images in a target page service session link, the information bearing capacity is rich, and the method can comprise all target page attention information in the target page service session link, so that mining analysis of page attention is facilitated; in addition, in view of the fact that the prior user page interaction behavior image applied by each push interest recognition network in the debugging period can comprise all interest details of target page attention matters, the debugging precision of each push interest recognition network is ensured, and the active debugging examples and the passive debugging examples of different push interest recognition network applications are different, so that each push interest recognition network completing the debugging can recognize the target page attention matters from different ideas, the push interest item recognition views of the target page attention matters can be accurately, comprehensively and quickly determined, and the push interest mining quality of page service users aiming at the push interest recognition processing is further improved.
In the embodiment of the invention, the target user page interactive behavior images are identified by utilizing a plurality of push interest identification networks in different ideas, the advantages of each push interest identification network are synthesized as much as possible, and the push interest identification networks for carrying out push interest identification through the commonality metric value among the user page interactive behavior images are utilized, so that the performance of the push interest identification networks for analyzing target page attention matters can be further enhanced.
In some possible examples, steps 41-44 may also be included.
Step 41: and when the target page attention items exist in the page service user to be subjected to the push interest identification processing, acquiring a target page service session link of the page service user to be subjected to the push interest identification processing.
When the target page attention matters exist in the page service users to be subjected to the push interest identification processing, the target page attention matters can be shared. Therefore, when it is determined that the target page attention item exists in the page service user to be subjected to the push interest identification process, it is required to analyze the page attention item of other page service users in the target page service session link where the page service user to be subjected to the push interest identification process is located, so as to determine whether the page attention item is sharable.
Step 42: and identifying the user page interaction behavior images of a plurality of session participants in the target page service session link based on each push interest identification network.
Under some exemplary design ideas, the user page interaction behavior images of a plurality of session participants in a target page service session are identified by adopting the debugged first push interest identification network, the second push interest identification network and the third push interest identification network, and whether the session participants have target page concerns or not is determined based on confidence coefficients of the existence of the target page concerns of each session participant.
Step 43: and determining the number of target page attention matters existing in the target page service session links in the set processing period based on the push interest item identification viewpoint.
In some cases, if, in a set processing period in the same page service session, there are a plurality of session participants that have the same target page attention as the page service user to be subjected to the push interest identification processing, it is indicated that the page attention has been shared in the target page service session corresponding to the page service user to be subjected to the push interest identification processing. Therefore, in step 43, the number of target page concerns is counted based on the push interest item identification points of the several session participants in the target page service session.
Step 44: and when the determined number exceeds the set limit value, determining the target page service session link as an active link of the target page attention item.
When the number of the target page attention items exceeds the set limit value in the implementation step 44, the sharable page attention items are characterized, the target page service session links can be determined to be active links of the target page attention items, and then group portrait analysis processing of the target page attention items can be performed for the target page service session links, so that a decision basis is provided for the subsequent batch big data pushing.
It will be appreciated that before implementing steps 1-3, the debugging process may be performed on different push interest recognition networks, in other words, the first push interest recognition network, the second push interest recognition network, and the third push interest recognition network may be respectively performed.
In some alternative embodiments, the first push interest recognition network is commissioned based on the steps of: obtaining a debugging example binary group set from the prior user page interaction behavior image; and performing repeated cyclic debugging on the first push interest recognition network to be debugged according to the debugging example binary group set, and outputting the first push interest recognition network with the completion of debugging.
Wherein, in each cycle debugging process, the following steps are implemented: inputting a plurality of selected debugging example tuples into the first pushing interest recognition network to be debugged, and respectively obtaining a debugging example commonality index between two debugging examples contained in each debugging example tuple; based on the obtained debug example commonality indexes, respectively obtaining page attention item prediction viewpoints of corresponding debug example doublets; and determining a push interest identification cost variable according to the page attention item authentication viewpoint and the page attention item prediction viewpoint corresponding to each of the plurality of debugging example tuples, and performing network improvement according to the push interest identification cost variable.
In the embodiment of the invention, the prior user page interaction behavior image can be understood as a user page interaction behavior image sample, the debugging example doublet can be understood as a debugging sample pair, and the debugging example doublet set can be understood as a set of the debugging sample pair. Further, the cyclic debugging can be understood as iterative training, the debugging example commonality index can be understood as sample similarity, the page attention item prediction viewpoint can be understood as a prediction label of the page attention item, the push interest recognition cost variable can be understood as a training loss value determined from the push interest recognition dimension, and the cyclic debugging of the first push interest recognition network is performed on the basis of the training loss value, so that the optimization of the first push interest recognition network can be continuously realized, and the robustness of the first push interest recognition network is improved.
In other alternative embodiments, the second push interest recognition network may be obtained based on the steps of: obtaining a debugging example set from the prior user page interactive behavior image, wherein the debugging example set comprises a plurality of positive debugging examples and a plurality of negative debugging examples; predicting network branches according to the set page attention items, and respectively implementing the following steps: according to the debug example set, obtaining respective page attention item prediction viewpoints of a plurality of debug examples, and determining a local viewpoint prediction cost variable according to respective page attention item authentication viewpoints and page attention item prediction viewpoints corresponding to the plurality of debug examples; aggregating a plurality of local viewpoint prediction cost variables, and selecting at least two target network branches from the plurality of page attention item prediction network branches based on the target cost variables after aggregation; integrating a selected minimum of two target network branches into the second push interest identification network.
In the debugging process for the second push interest identification network, the positive debugging example may be a positive sample and the negative debugging example may be a negative sample. The page attention item authentication view point and the page attention item prediction view point correspond to real page attention item labels and predicted page attention item labels respectively, based on the real page attention item labels and the predicted page attention item labels, local view point prediction cost variables (namely local prediction loss values) corresponding to page attention item prediction network branches can be determined, on the basis, a plurality of local view point prediction cost variables are subjected to weighting processing (weight values can be adjusted according to actual requirements), so that target cost variables of a global layer are obtained, target network branches are selected through the target cost variables, and then a second pushing interest identification network is further integrated. Therefore, the performance of the second pushing interest identification network can be guaranteed to the greatest extent by independently training and debugging the page interest prediction network branches and screening the page interest prediction network branches by combining the target cost variables which are aggregated.
In yet other alternative embodiments, the third push interest recognition network is commissioned based on the following ideas: obtaining a debugging example set from the prior user page interactive behavior image, wherein the debugging example set comprises a plurality of positive debugging examples and a plurality of negative debugging examples; according to the debugging example set, performing multiple times of cyclic debugging on a third pushing interest recognition network to be debugged, and outputting the third pushing interest recognition network with the completion of debugging, wherein in each cyclic debugging process, the following steps are implemented: according to the third pushing interest recognition network to be debugged, obtaining page attention description characteristics of a plurality of debugging examples; fusing and sizing the page focus description features of the plurality of debugging examples to obtain fused current page focus description features, and aggregating the current page focus description features with annotation features of the plurality of debugging examples; based on the combined characteristics of the completion of aggregation, determining a push interest identification cost variable, and carrying out network improvement according to the push interest identification cost variable.
In embodiments of the present invention, a debug instance may be understood as a debug sample. It should be appreciated that the debug samples/training samples of different push interest identification networks may be adaptively selected according to requirements, for example, the same samples may be selected, or different samples may be selected, which is not limited herein. Further, the page focus description feature is used to characterize the page behavior feature of the debug instance (i.e., the interactive behavior vector that a user may have potential interest in characterizing the user during a page interaction). In addition, the page attention description feature is subjected to splicing processing and dimension processing, so that the current page attention description feature (namely, the low-dimension page attention description feature after splicing) can be obtained, and then the current page attention description feature and the annotation feature (tag feature or priori feature) are subjected to fusion processing, so that the combined feature (mixed feature) for completing aggregation is obtained. In this way, the push interest identification cost variable may be determined by completing the aggregated joint feature, thereby enabling network improvement of the third push interest identification network.
Therefore, the debugging processing for the first push interest recognition network, the second push interest recognition network and the third push interest recognition network is realized based on different debugging ideas, so that the performance difference between each push interest recognition network can be ensured, and the push interest item recognition view mining of the target page attention item can be performed based on different angles.
In some independent embodiments, on the premise that the push interest item identification viewpoint characterizes the page service user to be subjected to push interest identification processing as the target page attention item, the method further comprises: determining push content of the page service user to be subjected to push interest identification processing according to the target page attention item, determining a push mode of the push content according to historical user behaviors of the page service user to be subjected to push interest identification processing, and generating an information push strategy of the page service user to be subjected to push interest identification processing by combining the push content and the push mode.
It can be appreciated that when determining the information push policy, the information push policy is implemented based on two dimensions, the first dimension focuses on push content of a page service user to be subjected to push interest identification processing, the push content can be determined through target page concerns, the second dimension focuses on push mode of the push content, and the push mode is determined based on historical user behaviors, so that the generated information push policy can comprehensively consider matching of the push content and the push mode to ensure accuracy and scene suitability of information push later, invalid push or repeated push caused by mismatching of the push content or the push mode is avoided, and maximum utilization of push resources (such as software and hardware resources) can be realized.
Further, there is also provided a readable storage medium having stored thereon a program which when executed by a processor implements the above-described method.
It will be clear to those skilled in the art that, for convenience and brevity of description, specific working procedures of the above-described system and apparatus may refer to corresponding procedures in the foregoing method embodiments, which are not repeated herein. In the several embodiments provided in the present invention, it should be understood that the disclosed system, apparatus and method may be implemented in other manners. The above-described apparatus embodiments are merely illustrative, for example, the division of the units is merely a logical function division, and there may be other manners of division in actual implementation, and for example, multiple units or components may be combined or integrated into another system, or some features may be omitted, or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be through some communication interface, device or unit indirect coupling or communication connection, which may be in electrical, mechanical or other form.
Claims (8)
1. A method for identifying a user interest, applied to a user interest identification device, the method comprising:
Obtaining a target user page interactive behavior image of a page service user to be subjected to push interest identification processing, and obtaining a push interest identification network set, wherein the push interest identification network set comprises a plurality of push interest identification networks for completing debugging;
mining corresponding interactive behavior state representation knowledge sets from the target user page interactive behavior images through each push interest recognition network, and obtaining confidence coefficients of target page concerns of the page service users to be subjected to push interest recognition processing, which are generated by each push interest recognition network, according to the interactive behavior state representation knowledge sets; the method comprises the steps that positive debugging examples and negative debugging examples of all pushing interest identification network applications are priori user page interactive behavior images in a target page service session link, and the positive debugging examples and the negative debugging examples of different pushing interest identification network applications are different;
determining that the page service user to be subjected to the push interest identification processing has a push interest item identification view of the target page attention item according to the confidence coefficient generated by each push interest identification network;
Wherein each push interest recognition network comprises at least one of the following:
the first push interest recognition network is used for carrying out push interest recognition through a commonality measurement value between the interactive behavior images of the user pages;
the second push interest recognition network carries out push interest recognition through a plurality of confidence coefficient predicted values corresponding to the user page interaction behavior image; wherein each confidence coefficient predictor is obtained by a target network branch in the second push interest identification network;
the third push interest recognition network carries out push interest recognition through linkage interaction behavior state characterization knowledge corresponding to the user page interaction behavior image;
when the push interest recognition networks are the first push interest recognition networks, mining corresponding interaction behavior state representation knowledge sets from the target user page interaction behavior images through the push interest recognition networks, and obtaining confidence coefficients of target page concerns of the page service users to be subjected to push interest recognition processing, which are generated by the push interest recognition networks, according to the interaction behavior state representation knowledge sets, wherein the confidence coefficients comprise:
According to the first push interest identification network, the following steps are implemented:
screening an auxiliary user page interaction behavior image from the prior user page interaction behavior image, and mining a corresponding auxiliary interaction behavior state representation knowledge set from the auxiliary user page interaction behavior image;
excavating a corresponding first target interaction behavior state representation knowledge set from the target user page interaction behavior image;
determining a push interest commonality score between the target user page interaction behavior image and the auxiliary user page interaction behavior image based on the first target interaction behavior state representation knowledge set and the auxiliary interaction behavior state representation knowledge set;
generating a first confidence coefficient of target page attention matters existing in the page service user to be subjected to push interest identification processing based on the push interest commonality scores;
wherein, still include:
when the auxiliary user page interaction behavior image is a user page interaction behavior image with a relationship of the target page attention item, a first quantification relationship exists between the push interest commonality score and the first confidence coefficient;
When the auxiliary user page interaction behavior image is a user page interaction behavior image with a relationship of non-target page attention items, a second quantification relationship exists between the push interest commonality score and the first confidence coefficient;
wherein:
the auxiliary user page interactive behavior image is a reference user page interactive behavior image; the auxiliary user page interaction behavior image is a user page interaction behavior image with a relationship of target page attention matters or a user page interaction behavior image with a relationship of non-target page attention matters;
the push interest commonality score is the similarity of the push interest features.
2. The method of claim 1, wherein the first push interest recognition network is commissioned based on the following ideas:
obtaining a debugging example binary group set from the prior user page interaction behavior image;
according to the debugging example binary group set, performing repeated cyclic debugging on a first pushing interest recognition network to be debugged, and outputting the first pushing interest recognition network with the completion of the debugging, wherein in the process of each cyclic debugging, the following steps are implemented:
inputting a plurality of selected debugging example tuples into the first pushing interest recognition network to be debugged, and respectively obtaining a debugging example commonality index between two debugging examples contained in each debugging example tuple;
Based on the obtained debug example commonality indexes, respectively obtaining page attention item prediction viewpoints of corresponding debug example doublets;
determining a push interest identification cost variable according to the page attention item authentication viewpoint and the page attention item prediction viewpoint corresponding to each of the plurality of debugging example tuples, and performing network improvement according to the push interest identification cost variable;
wherein the push interest identification cost variable is a training loss value determined from the push interest identification dimension.
3. The method according to claim 1, wherein when the push interest recognition networks are the second push interest recognition networks, mining, by the push interest recognition networks, a corresponding interaction behavior state characterization knowledge set from the target user page interaction behavior image, and obtaining, according to the interaction behavior state characterization knowledge set, a confidence coefficient of a target page interest item existing in the page service user to be subjected to push interest recognition processing generated by the push interest recognition networks, includes:
respectively mining a second target interaction behavior state representation knowledge set of the target page attention items according to a plurality of target network branches contained in the second push interest identification network;
Based on the second target interaction behavior state representation knowledge set corresponding to each of the plurality of target network branches, obtaining a confidence coefficient predicted value of the target user page interaction behavior image generated by the corresponding target network branch, wherein the confidence coefficient predicted value belongs to target page attention matters;
optimizing a plurality of confidence coefficient predicted values based on the importance degree of the set page attention items corresponding to the target network branches respectively, and obtaining a second confidence coefficient of the target page attention items of the page service user to be subjected to the push interest identification processing, which is generated by the second push interest identification network;
wherein the second push interest recognition network is obtained based on the steps of:
obtaining a debugging example set from the prior user page interactive behavior image, wherein the debugging example set comprises a plurality of positive debugging examples and a plurality of negative debugging examples;
predicting network branches according to the set page attention items, and respectively implementing the following steps:
according to the debug example set, obtaining respective page attention item prediction viewpoints of a plurality of debug examples, and determining a local viewpoint prediction cost variable according to respective page attention item authentication viewpoints and page attention item prediction viewpoints corresponding to the plurality of debug examples;
Aggregating a plurality of local viewpoint prediction cost variables, and selecting at least two target network branches from the plurality of page attention item prediction network branches based on the target cost variables after aggregation;
integrating at least two selected target network branches into the second push interest identification network;
the local viewpoint prediction cost variable is a local prediction loss value.
4. The method according to claim 1, wherein when the push interest recognition networks are the third push interest recognition networks, mining, by the push interest recognition networks, a corresponding interaction behavior state characterization knowledge set from the target user page interaction behavior image, and obtaining, according to the interaction behavior state characterization knowledge set, a confidence coefficient of a target page interest item existing in the page service user to be subjected to push interest recognition processing generated by the push interest recognition networks, includes:
according to the third pushing interest recognition network, a corresponding third target interaction behavior state representation knowledge set is obtained from the target user page interaction behavior image through excavation;
Performing aggregation operation on each interactive behavior state representation knowledge in the third target interactive behavior state representation knowledge set, and obtaining a third confidence coefficient of target page concern items of the page service user to be subjected to push interest identification processing, which is generated by the third push interest identification network, based on the aggregated interactive behavior state representation knowledge;
the third push interest recognition network is obtained based on the following idea:
obtaining a debugging example set from the prior user page interactive behavior image, wherein the debugging example set comprises a plurality of positive debugging examples and a plurality of negative debugging examples;
according to the debugging example set, performing multiple times of cyclic debugging on a third pushing interest recognition network to be debugged, and outputting the third pushing interest recognition network with the completion of debugging, wherein in each cyclic debugging process, the following steps are implemented:
according to the third pushing interest recognition network to be debugged, obtaining page attention description characteristics of a plurality of debugging examples;
fusing and sizing the page focus description features of the plurality of debugging examples to obtain fused current page focus description features, and aggregating the current page focus description features with annotation features of the plurality of debugging examples;
Based on the combined characteristics of the completion of aggregation, determining a push interest identification cost variable, and carrying out network improvement according to the push interest identification cost variable.
5. The method of claim 1, wherein when the respective push interest recognition networks include a first push interest recognition network, a second push interest recognition network, and a third push interest recognition network, determining a push interest item recognition perspective of the target page interest based on the respective generated confidence coefficients of the respective push interest recognition networks comprises:
when a first confidence coefficient determined by a first push interest identification network is larger than a set first page attention item limit value, determining that a push interest item identification view of the target page attention item exists in the page service user to be subjected to push interest identification processing according to the first confidence coefficient;
when the minimum value of the second confidence coefficient determined by the second push interest recognition network and the minimum value of the third confidence coefficient determined by the third push interest recognition network are both larger than the set second page attention item limit value, determining that the page service user to be subjected to push interest recognition processing has a push interest item recognition view of the target page attention item according to the maximum value of the second confidence coefficient and the third confidence coefficient;
And when the first confidence coefficient is not greater than the first page attention item limit value and the minimum value of the second confidence coefficient and the third confidence coefficient is smaller than or equal to the second page attention item limit value, determining that the page service user to be subjected to push interest identification processing has a push interest item identification view of the target page attention item according to the minimum value of the second confidence coefficient and the third confidence coefficient.
6. The method of claim 5, wherein when it is determined that the target page attention item exists for the page service user to be subjected to the push interest identification process, the method further comprises:
obtaining a target page service session link of the page service user to be subjected to push interest identification processing;
identifying user page interaction behavior images of a plurality of session participants in the target page service session link according to the push interest identification networks;
determining the number of target page attention matters in the target page service session links in a set processing period based on a push interest item identification viewpoint;
And when the number exceeds a set limit value, determining the target page service session link as an active link of the target page attention item.
7. A user interest recognition device, comprising a processor and a memory; the processor being communicatively connected to the memory, the processor being adapted to read a computer program from the memory and execute it to carry out the method of any of the preceding claims 1-6.
8. A computer readable storage medium, characterized in that it has stored thereon a computer program, which, when run, implements the method of any of the preceding claims 1-6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311234324.7A CN116975455B (en) | 2023-09-24 | 2023-09-24 | User interest recognition method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311234324.7A CN116975455B (en) | 2023-09-24 | 2023-09-24 | User interest recognition method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116975455A CN116975455A (en) | 2023-10-31 |
| CN116975455B true CN116975455B (en) | 2023-12-22 |
Family
ID=88473448
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311234324.7A Active CN116975455B (en) | 2023-09-24 | 2023-09-24 | User interest recognition method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116975455B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114840486A (en) * | 2022-06-28 | 2022-08-02 | 广州趣米网络科技有限公司 | User behavior data acquisition method and system and cloud platform |
| CN115438267A (en) * | 2022-09-27 | 2022-12-06 | 黄浩 | Portrait analysis processing method and system based on big data |
| CN116089727A (en) * | 2023-03-09 | 2023-05-09 | 陆路 | Information analysis method and system based on user portrait mining |
| CN116186372A (en) * | 2023-03-07 | 2023-05-30 | 南京大学 | Bibliographic system capable of providing personalized service |
| CN116362782A (en) * | 2022-12-31 | 2023-06-30 | 广州东朝智能科技有限公司 | User interest point identification method and system based on big data analysis |
-
2023
- 2023-09-24 CN CN202311234324.7A patent/CN116975455B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114840486A (en) * | 2022-06-28 | 2022-08-02 | 广州趣米网络科技有限公司 | User behavior data acquisition method and system and cloud platform |
| CN115438267A (en) * | 2022-09-27 | 2022-12-06 | 黄浩 | Portrait analysis processing method and system based on big data |
| CN116362782A (en) * | 2022-12-31 | 2023-06-30 | 广州东朝智能科技有限公司 | User interest point identification method and system based on big data analysis |
| CN116186372A (en) * | 2023-03-07 | 2023-05-30 | 南京大学 | Bibliographic system capable of providing personalized service |
| CN116089727A (en) * | 2023-03-09 | 2023-05-09 | 陆路 | Information analysis method and system based on user portrait mining |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116975455A (en) | 2023-10-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Xu et al. | Adapting user preference to online feedback in multi-round conversational recommendation | |
| Song et al. | Single image dehazing using ranking convolutional neural network | |
| CN116049397B (en) | Sensitive information discovery and automatic classification method based on multi-mode fusion | |
| CN113762280A (en) | Image category identification method, device and medium | |
| CN114048387B (en) | Content recommendation method based on big data and AI prediction and artificial intelligence cloud system | |
| CN114041165A (en) | Video similarity detection method, device and equipment | |
| US20230049817A1 (en) | Performance-adaptive sampling strategy towards fast and accurate graph neural networks | |
| Yang et al. | Deep feature importance awareness based no-reference image quality prediction | |
| Li et al. | Meta parsing networks: Towards generalized few-shot scene parsing with adaptive metric learning | |
| CN112100442A (en) | User tendency recognition method, device, equipment and storage medium | |
| JP2021192286A (en) | Model training, image processing method and device, storage medium, and program product | |
| Srivastava et al. | Fraud detection in the distributed graph database | |
| CN115238191A (en) | Object recommendation method and device | |
| CN114678137A (en) | Big data mining method and server based on digital medical treatment | |
| Shu et al. | Semi-supervised adversarial learning for attribute-aware photo aesthetic assessment | |
| CN114398973A (en) | Media content label identification method, device, equipment and storage medium | |
| Gan et al. | A multi-behavior recommendation method exploring the preference differences among various behaviors | |
| CN111597361B (en) | Multimedia data processing method, device, storage medium and equipment | |
| CN111582448B (en) | Weight training method and device, computer equipment and storage medium | |
| CN116975455B (en) | User interest recognition method and device | |
| CN113407837A (en) | Intelligent medical big data processing method based on artificial intelligence and intelligent medical system | |
| CN114625612A (en) | User behavior analysis method and service system based on big data office | |
| CN114168966B (en) | Big data analysis-based security protection upgrade mining method and information security system | |
| CN114647787A (en) | User personalized recommendation method based on multi-modal data | |
| CN114625961A (en) | Intelligent online service pushing method applied to big data and big data server |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |