CN116362782A - User interest point identification method and system based on big data analysis - Google Patents
User interest point identification method and system based on big data analysis Download PDFInfo
- Publication number
- CN116362782A CN116362782A CN202211740018.6A CN202211740018A CN116362782A CN 116362782 A CN116362782 A CN 116362782A CN 202211740018 A CN202211740018 A CN 202211740018A CN 116362782 A CN116362782 A CN 116362782A
- Authority
- CN
- China
- Prior art keywords
- template
- debugging
- interest point
- error value
- target
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 84
- 238000007405 data analysis Methods 0.000 title claims abstract description 14
- 238000004458 analytical method Methods 0.000 claims abstract description 256
- 230000006399 behavior Effects 0.000 claims abstract description 34
- 239000013598 vector Substances 0.000 claims description 251
- 238000012512 characterization method Methods 0.000 claims description 178
- 230000006870 function Effects 0.000 claims description 20
- 238000005065 mining Methods 0.000 claims description 19
- 238000000605 extraction Methods 0.000 claims description 18
- 238000004140 cleaning Methods 0.000 claims description 13
- 230000001939 inductive effect Effects 0.000 claims description 12
- 238000013500 data storage Methods 0.000 claims description 6
- 238000007670 refining Methods 0.000 claims description 6
- 238000004590 computer program Methods 0.000 claims description 4
- 230000014759 maintenance of location Effects 0.000 claims description 4
- 238000012216 screening Methods 0.000 claims description 3
- 230000000875 corresponding effect Effects 0.000 description 237
- 230000008569 process Effects 0.000 description 22
- 238000010586 diagram Methods 0.000 description 9
- 238000004364 calculation method Methods 0.000 description 6
- 238000004891 communication Methods 0.000 description 5
- 238000012545 processing Methods 0.000 description 5
- 238000013140 knowledge distillation Methods 0.000 description 4
- 230000009471 action Effects 0.000 description 3
- 239000006185 dispersion Substances 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 238000003860 storage Methods 0.000 description 3
- 238000013527 convolutional neural network Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 230000006698 induction Effects 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000003542 behavioural effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000004821 distillation Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000012552 review Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/245—Query processing
- G06F16/2458—Special types of queries, e.g. statistical queries, fuzzy queries or distributed queries
- G06F16/2462—Approximate or statistical queries
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/245—Query processing
- G06F16/2458—Special types of queries, e.g. statistical queries, fuzzy queries or distributed queries
- G06F16/2465—Query processing support for facilitating data mining operations in structured databases
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2216/00—Indexing scheme relating to additional aspects of information retrieval not explicitly covered by G06F16/00 and subgroups
- G06F2216/03—Data mining
-
- 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)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Development Economics (AREA)
- Strategic Management (AREA)
- Finance (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Probability & Statistics with Applications (AREA)
- Accounting & Taxation (AREA)
- Databases & Information Systems (AREA)
- Mathematical Physics (AREA)
- Entrepreneurship & Innovation (AREA)
- Fuzzy Systems (AREA)
- Game Theory and Decision Science (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Economics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Debugging And Monitoring (AREA)
Abstract
According to the user interest point identification method and system based on big data analysis, the user interest point analysis network is used for analyzing and identifying the user exhibition behavior data to obtain the interest points of the user on the exhibition, wherein the adopted user interest point analysis network is based on continuous learning, debugging and acquisition, knowledge of one task is applied to the other task, meanwhile, knowledge capacity of the previous task is reserved, and accordingly the universality of newly introduced interest points and existing interest points can be enhanced on the premise that knowledge acquired by the debugged user interest point analysis network is reserved by the user interest point analysis network, and accuracy of interest point analysis is improved when the interest point analysis is carried out on the basis of the user interest point analysis network.
Description
Technical Field
The application relates to the field of artificial intelligence and data processing, in particular to a user interest point identification method and system based on big data analysis.
Background
With the development of Internet and big data technology, internet and big data are permeated into every corner in life, a large amount of behavior big data and basic static data of a target user can be acquired on the premise of legal compliance through the Internet and big data technology, and a platform party can be helped to perform legal and reasonable analysis and understanding on the user through analyzing the big data, so that the promotion of related services is promoted. Currently, the method has been widely applied to Internet platforms such as electronic commerce, short video, electronic reading and the like. There is no intervention of related technology in the exhibition field, or simple data statistics is performed to summarize feedback information of exhibition users, and obviously, the method is rough and cannot accurately analyze and identify interest points of the users. In addition, the interest points of the user are dynamically changed, and are analyzed based on the existing interest point analysis mode, so that the method cannot be well suitable for analysis and identification after new interest points are generated. It should be noted that the foregoing description of the background art is only for convenience of understanding the background of the present application, and does not generate any limitation or technical suggestion for the technical solution of the present application.
Disclosure of Invention
The invention aims to provide a user interest point identification method and system based on big data analysis so as to improve the problems.
The implementation manner of the embodiment of the application is as follows:
in a first aspect, an embodiment of the present application provides a method for identifying a user interest point based on big data analysis, which is applied to a system for identifying a user interest point, and the method includes:
responding to a user analysis instruction, and acquiring user exhibition behavior data in a preset behavior data storage space;
loading the user exhibition behavior data into a user interest point analysis network, and analyzing the interest points of the user exhibition behavior data based on the user interest point analysis network;
the user interest point analysis network is obtained by adjusting network parameters of a user interest point analysis network to be debugged according to a first target error value, a second target error value and an identification error value, wherein the first target error value is obtained by carrying out error determination according to a first template characterization vector and a second template characterization vector of the same debugging template, the second target error value is obtained by carrying out error determination according to second template characterization vectors respectively corresponding to debugging template groups with different interest points, the first template characterization vector of the debugging template is obtained by extracting based on the debugged user interest point analysis network, the second template characterization vector of the debugging template is obtained by extracting based on the user interest point analysis network to be debugged, the identification error value is obtained by carrying out interest point analysis on the second template characterization vector of the debugging template according to the user interest point analysis network to be debugged, and the debugging template is contained in a debugging template set which contains templates with determined interest points corresponding to the debugged user interest point analysis network and debugging templates with new introduced interest points;
And obtaining an interest point analysis result corresponding to the user exhibition behavior data output by the user interest point analysis network.
Further, the method further comprises a step of debugging the user interest point analysis network, comprising:
acquiring a debugging template set, wherein the debugging template set comprises a debugging template with determined interest points corresponding to a debugged user interest point analysis network and a debugging template with newly introduced interest points;
respectively mining first template characterization vectors corresponding to each debugging template in the debugging template set based on the debugged user interest point analysis network;
respectively mining second template characterization vectors corresponding to the debugging templates based on a user interest point analysis network to be debugged, carrying out interest point analysis according to the second template characterization vectors, and determining identification error values according to the obtained first interest point analysis results;
performing error determination according to a first template characterization vector and a second template characterization vector of the same debugging template to obtain a first target error value, and performing error determination according to second template characterization vectors respectively corresponding to the debugging template groups with different interest points to obtain a second target error value;
And adjusting network parameter values of the interest point analysis network of the user to be debugged according to the first target error value, the second target error value and the identification error value, performing iterative debugging, and obtaining a target user interest point analysis network if the debugging cut-off requirement is met, wherein the target user interest point analysis network is configured to analyze and identify the determined interest point and the newly introduced interest point.
Further, the error determination is performed according to the second template characterization vectors respectively corresponding to the debug template groups with different interest points, so as to obtain a second target error value, which includes:
respectively constructing each debugging template and each debugging template in the debugging template set into a debugging template group to obtain a plurality of debugging template groups;
determining a common coefficient according to the second template characterization vectors respectively corresponding to the debugging template groups, and acquiring sub-error values respectively corresponding to the debugging template groups according to the determined common coefficient;
and inducing sub-error values corresponding to the target debugging template group to obtain a second target error value, wherein the target debugging template group is a debugging template group of the debugging templates with different interest points.
Further, the obtaining the sub-error value corresponding to each debug template group according to the determined commonality coefficient includes:
For the common coefficient corresponding to each debugging template, the common coefficient is subjected to difference with a preset common coefficient to obtain a target coefficient difference;
when the common coefficient is smaller than the preset common coefficient, determining the minimum error value as a sub-error value corresponding to the debugging template;
when the common coefficient is not smaller than the preset common coefficient, determining the target coefficient difference as a sub-error value corresponding to the debugging template;
the sub-error value corresponding to the generalized target debugging template group is obtained, and the second target error value is obtained, including:
for each debugging template group, obtaining interest point annotation information corresponding to a debugging template included in the debugging template group, and calculating an illustrative result corresponding to the debugging template group according to the interest point annotation information; the method comprises the steps that an illustrative result is determined based on an illustrative function, when interest point annotation information corresponding to a debugging template included in a debugging template group is the same, the illustrative result determined based on the illustrative function is a first character, and when interest point annotation information corresponding to the debugging template included in the debugging template group is the same, the illustrative result determined based on the illustrative function is a second character;
when the obtained illustrative result is determined to be a first character, carrying out a retention operation on the sub-error value corresponding to the debugging template group to obtain a target sub-error value corresponding to the debugging template group, and when the obtained illustrative result is determined to be a second character, carrying out a cleaning operation on the sub-error value corresponding to the debugging template group to obtain a target sub-error value corresponding to the debugging template group;
And inducing target sub-error values corresponding to the debugging template groups respectively to obtain second target error values.
Further, before the adjusting the network parameter of the to-be-debugged user interest point analysis network according to the first target error value, the second target error value and the identification error value and performing iterative debugging, the method further includes:
performing error determination according to the first template characterization vector and the second template characterization vector corresponding to the debugging template groups with different interest points to obtain a third target error value;
the adjusting the network parameter of the user interest point analysis network to be debugged according to the first target error value, the second target error value and the identification error value, and performing iterative debugging, including:
inducing the first target error value, the second target error value, the third target error value and the identification error value to obtain an induced error value;
and adjusting the network parameter of the interest point analysis network of the user to be debugged according to the induced error value and performing iterative debugging.
Further, the error determination is performed according to the first template characterization vector and the second template characterization vector of the same debug template, to obtain a first target error value, including:
Determining a common coefficient according to a first template characterization vector and a second template characterization vector which correspond to each debugging template respectively, and acquiring respective sub-error values of each debugging template according to the determined common coefficient; the sub-error value of the debugging template is reversely related to the commonality coefficient corresponding to the debugging template;
summarizing respective sub-error values of each debugging template to obtain a first target error value;
the determined interest point comprises a plurality of interest points, and the debug template set is constructed by the steps of:
acquiring a basic template set in a preset data space; the basic template set comprises a reference debugging template corresponding to the determined interest point, and the scattering condition of the reference debugging template corresponding to the determined interest point in the characterization vector value domain corresponding to the determined interest point is associated with the scattering condition of the basic template set corresponding to the determined interest point in the characterization vector value domain;
acquiring a new introduced template set corresponding to the new introduced interest point, and constructing a debugging template set according to the new introduced template set and the basic template set;
before the basic template set is acquired in the preset data space, the method further comprises the following steps:
Acquiring a basic template set corresponding to each determined interest point respectively; the basic template set comprises an analyzed debugging template corresponding to the debugged user interest point analysis network;
grouping template characterization vectors corresponding to each analyzed and debugged template in the basic template set to obtain a plurality of grouping results corresponding to the determined interest points;
screening template characterization vectors with vector distances meeting the preset distance requirement from the respective grouping results, and determining an analyzed debugging template corresponding to the screened template characterization vectors as a reference debugging template corresponding to the determined interest point.
Further, the grouping of the template characterization vectors corresponding to each analyzed and debugged template in the basic template set to obtain a plurality of grouping results corresponding to the determined interest point includes:
determining a first interest point statistics value corresponding to the determined interest point, and determining a target interest point statistics value during grouping according to the first interest point statistics value and the data space size of the preset data space;
Grouping template characterization vectors respectively corresponding to all analyzed and debugged templates in the basic template set according to the target interest point statistics value to obtain a plurality of grouping results corresponding to the determined interest point and matched with the target interest point statistics value.
Further, the newly introduced point of interest includes a plurality of points of interest, and after obtaining the target user point of interest analysis network, the method further includes:
determining a second interest point statistics value corresponding to the newly introduced interest point, and cleaning a plurality of reference debugging templates in the basic template set in the preset data space according to the second interest point statistics value;
determining a reference debugging template corresponding to the new introduced interest point in a new introduced template set corresponding to the new introduced interest point;
storing the reference debugging template corresponding to the newly introduced interest point into the preset data space;
the cleaning the plurality of reference debugging templates in the basic template set in the preset data space according to the second interest point statistics comprises the following steps:
determining the target template quantity of the templates to be cleaned corresponding to each determined interest point according to the second interest point statistics;
for each determined interest point, determining a representative characterization vector of a characterization vector value field corresponding to the determined interest point according to an analyzed debugging template corresponding to the determined interest point;
Respectively obtaining vector distances between template characterization vectors and the representative characterization vectors respectively corresponding to the reference debugging templates;
and cleaning the reference template characterization vectors matched with the number of the target templates from the reference template characterization vectors corresponding to the determined interest points according to the vector distances respectively corresponding to the reference debugging templates.
Further, the first interest point analysis result comprises a first reasoning confidence degree corresponding to the determined interest point;
before the adjusting the network parameter of the user interest point analysis network to be debugged according to the first target error value, the second target error value and the identification error value and performing iterative debugging, the method further comprises:
based on the debugged user interest point analysis network, interest point analysis is carried out according to the first template characterization vectors respectively corresponding to the debug templates, and respective second interest point analysis results of the debug templates are obtained; the second interest point analysis result comprises a second reasoning confidence degree corresponding to the determined interest point;
for each debugging template, carrying out standardization operation on the first reasoning confidence coefficient and the second reasoning confidence coefficient of the debugging template under the same refined variable to obtain a first target confidence coefficient corresponding to the first reasoning confidence coefficient and a second target confidence coefficient corresponding to the second reasoning confidence coefficient;
Determining a refining error value according to the first target confidence coefficient and the second target confidence coefficient to obtain a refining error value of the debugging template;
summarizing the extraction error values of each debugging template to obtain a target extraction error value;
the adjusting the network parameter of the user interest point analysis network to be debugged according to the first target error value, the second target error value and the identification error value, and performing iterative debugging, including:
inducing the first target error value, the second target error value, the target extraction error value and the identification error value to obtain an induced error value;
and adjusting the network parameter of the interest point analysis network of the user to be debugged according to the induced error value and performing iterative debugging.
In another aspect, embodiments of the present application provide a user point of interest analysis system comprising a processor and a memory storing a computer program for execution by the processor to implement the method above.
According to the user interest point identification method and system based on big data analysis, the user interest point analysis network adopted for analysis and identification of the user exhibition behavior data is obtained based on continuous learning debugging, knowledge of one task is applied to the other task, meanwhile, knowledge capacity of the previous task is reserved, and on the premise that knowledge obtained by the debugged user interest point analysis network is reserved, universality and generalization capacity of newly introduced interest points and existing interest points can be enhanced by the user interest point analysis network, so that accuracy of interest point analysis is improved when the interest point analysis is carried out based on the user interest point analysis network.
In the following description, other features will be partially set forth. Upon review of the ensuing disclosure and the accompanying figures, those skilled in the art will in part discover these features or will be able to ascertain them through production or use thereof. The features of the present application may be implemented and obtained by practicing or using the various aspects of the methods, tools, and combinations that are set forth in the detailed examples described below.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are needed in the embodiments will be briefly described below, it being understood that the following drawings only illustrate some embodiments of the present application and therefore should not be considered limiting the scope, and that other related drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
The methods, systems, and/or programs in the accompanying drawings will be described further in terms of exemplary embodiments. These exemplary embodiments will be described in detail with reference to the drawings. These exemplary embodiments are non-limiting exemplary embodiments, wherein reference numerals represent similar mechanisms throughout the several views of the drawings.
Fig. 1 is a schematic illustration of an application scenario shown according to some embodiments of the present application.
FIG. 2 is a schematic diagram of hardware and software components in a user point of interest identification system, according to some embodiments of the present application.
FIG. 3 is a flow chart illustrating a method of user point of interest identification based on big data analysis, according to some embodiments of the present application.
Fig. 4 is a schematic architecture diagram of a user interest point identification device according to an embodiment of the present application.
Detailed Description
In order to better understand the technical solutions described above, the following detailed description of the technical solutions of the present application is provided through the accompanying drawings and specific embodiments, and it should be understood that the specific features of the embodiments and embodiments of the present application are detailed descriptions of the technical solutions of the present application, and not limit the technical solutions of the present application, and the technical features of the embodiments and embodiments of the present application may be combined with each other without conflict.
In the following detailed description, numerous specific details are set forth by way of examples in order to provide a thorough understanding of the relevant teachings. However, it will be apparent to one skilled in the art that the present application may be practiced without these details. In other instances, well-known methods, procedures, systems, components, and/or circuits have been described at a relatively high-level, without detail, in order to avoid unnecessarily obscuring aspects of the present application.
These and other features, together with the functions, acts, and combinations of parts and economies of manufacture of the related elements of structure, all of which form part of this application, may become more apparent upon consideration of the following description with reference to the accompanying drawings. It is to be expressly understood, however, that the drawings are for the purpose of illustration and description only and are not intended as a definition of the limits of the application. It should be understood that the drawings are not to scale. It is to be expressly understood, however, that the drawings are for the purpose of illustration and description only and are not intended as a definition of the limits of the application. It should be understood that the figures are not to scale.
The flowcharts are used in this application to describe implementations performed by systems according to embodiments of the present application. It should be clearly understood that the execution of the flowcharts may be performed out of order. Rather, these implementations may be performed in reverse order or concurrently. Additionally, at least one other execution may be added to the flowchart. One or more of the executions may be deleted from the flowchart.
Fig. 1 is a schematic diagram of an application scenario, according to some embodiments of the present application, in which a user point of interest identification system 100 and a client 300 are communicatively connected through a network 200.
In some embodiments, please refer to fig. 2, which is a schematic architecture diagram of the user interest point identification system 100, the user interest point identification system 100 includes a user interest point identification device 110, a memory 120, a processor 130 and a communication unit 140. The memory 120, the processor 130, and the communication unit 140 are electrically connected directly or indirectly to each other to realize data transmission or interaction. For example, the components may be electrically connected to each other via one or more communication buses or signal lines. The user point of interest identification means 110 comprises at least one software functional module which may be stored in the memory 120 in the form of software or firmware (firmware) or cured in an Operating System (OS) of the user point of interest identification system 100. The processor 130 is configured to execute executable modules stored in the memory 120, such as software functional modules and computer programs included in the user point of interest identification means 110. The user point of interest identification system 100 may be a server or a computer device. The Memory 120 may be, but is not limited to, a random access Memory (Random Access Memory, RAM), a Read Only Memory (ROM), a programmable Read Only Memory (Programmable Read-Only Memory, PROM), an erasable Read Only Memory (Erasable Programmable Read-Only Memory, EPROM), an electrically erasable Read Only Memory (Electric Erasable Programmable Read-Only Memory, EEPROM), etc. The memory 120 is used for storing a program, and the processor 130 executes the program after receiving an execution instruction. The communication unit 140 is used for establishing a communication connection between the user point of interest recognition system 100 and the front-end image pickup apparatus 200 through a network, and for transceiving data through the network. The processor may be an integrated circuit chip having signal processing capabilities. The processor may be a general-purpose processor, including a central processing unit (Central Processing Unit, CPU), a network processor (Network Processor, NP), etc.; but also Digital Signal Processors (DSPs)), application Specific Integrated Circuits (ASICs), field Programmable Gate Arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. The disclosed methods, steps, and logic blocks in the embodiments of the present invention may be implemented or performed. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
It is to be appreciated that the configuration shown in FIG. 2 is merely illustrative and that the user point of interest identification system 100 may also include more or fewer components than shown in FIG. 2 or have a different configuration than shown in FIG. 2. The components shown in fig. 2 may be implemented in hardware, software, or a combination thereof.
Fig. 3 is a flowchart of a method for identifying user interest points based on big data analysis according to some embodiments of the present application, which is applied to the user interest point identification system 100 in fig. 1, and may specifically include the following steps S10 to S30. On the basis of the following steps, alternative embodiments will be described, which should be understood as examples and should not be interpreted as essential features for implementing the present solution.
S10: and responding to the user analysis instruction, and acquiring user exhibition behavior data in a preset behavior data storage space.
In the embodiment of the application, the exhibition behavior data of the user is obtained by means of online software of the exhibition party, such as an exhibition applet, an exhibition app, an exhibition two-dimension code information collecting channel and the like, in the process of participating in the exhibition by the user, for example, related behavior operation is performed based on the exhibition application program carried by the client 300. The exhibition behavior data include, but are not limited to, reservation registration information, exhibit search data, exhibit resident data, exhibit evaluation data, exhibit attention data, etc., which are obtained by embedding a point identification (the type of the embedded point is not limited, and can be adapted to the analysis requirements) in advance, and the data are collected by using a preset collection tool (such as Flume, kafka, multiplexing, etc.), and output and stored in a preset behavior data storage space. And when receiving a user analysis instruction aiming at the target user, retrieving user exhibition behavior data corresponding to the target user from the behavior data storage space.
S20: and loading the user exhibition behavior data into a user interest point analysis network, and carrying out interest point analysis on the user exhibition behavior data based on the user interest point analysis network.
In the embodiment of the application, the user interest point analysis network is an artificial intelligent network established based on any feasible machine learning or deep learning network architecture, such as a Convolutional Neural Network (CNN), a multi-layer perceptron (MLP), a cyclic or Recurrent Neural Network (RNN), a long-short-term memory network (LSTM), a transducer, and the like. The method is obtained by adjusting network parameters of the user interest point analysis network to be debugged according to the first target error value, the second target error value and the identification error value.
The first target error value is obtained by carrying out error determination according to a first template characterization vector and a second template characterization vector of the same debugging template, the second target error value is obtained by carrying out error determination according to second template characterization vectors respectively corresponding to debugging template groups with different interest points, the first template characterization vector of the debugging template is obtained by extracting based on a debugged user interest point analysis network, the second template characterization vector of the debugging template is obtained by extracting based on a to-be-debugged user interest point analysis network, the identification error value is obtained according to an interest point analysis result generated by carrying out interest point analysis on the second template characterization vector of the debugging template according to the to-be-debugged user interest point analysis network, and the debugging template is contained in a debugging template set which contains the debugging templates with determined interest points corresponding to the debugged user interest point analysis network and the debugging templates with newly introduced interest points. The debugging process of the user point of interest analysis network will be described in detail later. The user interest point analysis network is used for carrying out interest point analysis on the user exhibition behavior data and identifying the interest points of the users implied in the user exhibition behavior data, such as the image of the exhibition intention exhibit.
S30: and obtaining an interest point analysis result corresponding to the user exhibition behavior data output by the user interest point analysis network.
The user interest point identification system acquires an interest point analysis result corresponding to the user exhibition behavior data output by the user interest point analysis network, and the interest point analysis result indicates the user interest point corresponding to the user exhibition behavior data.
Through the steps S10 to S30, the mining of the interest points of the exhibition behavior data of the user is completed, and the speed of the process is improved and the accuracy is high by means of the artificial intelligent network, and the specific debugging process of the interest point analysis network of the user is introduced below, which can comprise the following steps:
s100, obtaining a debugging template set.
The debugging template set comprises a debugging template with determined interest points corresponding to the debugged user interest point analysis network and a debugging template with newly introduced interest points. In this embodiment of the present application, the debug template set includes a plurality of debug templates, where the debug templates are exhibition behavioral data templates for debugging the user interest point analysis network, and the obtaining may be obtained from public, historical exhibition data. The debugging templates in the debugging template set correspond to different types, including determined interest points and newly introduced interest points corresponding to the debugged user interest point analysis network, wherein the debugged user interest point analysis network is a user interest point analysis network which is debugged in the past, the determined interest points corresponding to the debugged user interest point analysis network are interest point types which can be obtained by analysis of the debugged user interest point analysis network, the interest point types which can be obtained by analysis of the debugged user interest point analysis network are interest point types learned during debugging of the user interest point analysis network, and all the past known interest point types can be obtained by the debugged user interest point analysis network during debugging. The newly introduced interest point is the interest point type generated after the debugged user interest point analysis network is debugged. At least one of the determined interest point and the newly introduced interest point.
The debugging template set comprises a debugging template with determined interest points corresponding to the debugged user interest point analysis network and a debugging template with newly introduced interest points. The user interest point analysis system can debug the user interest point analysis network to be debugged according to the debugging template set so that the user interest point analysis network to be debugged can complete continuous learning on the basis of the debugged user interest point analysis network, and the user interest point analysis network capable of analyzing and identifying the determined interest points and the newly introduced interest points is obtained.
S200: and respectively mining first template characterization vectors corresponding to each debugging template in the debugging template set based on the debugged user interest point analysis network.
In this embodiment of the present application, the user interest point analysis system may load each debug template in the debug template set into the debugged user interest point analysis network, and the debugged user interest point analysis network may include a token vector mining module, where the token vector mining module may perform token vector mining on each debug template to obtain a first template token vector corresponding to each debug template. For example, the token vector mining module comprises a coding unit and a token vector mining unit, the user interest point analysis system loads the debugging template into the debugged user interest point analysis network, codes the debugging template based on the coding unit, and extracts the token vector based on the token vector mining unit to obtain the token vector.
S300: and respectively mining second template characterization vectors corresponding to each debugging template based on the interest point analysis network of the user to be debugged, carrying out interest point analysis according to each second template characterization vector, and determining an identification error value according to each obtained first interest point analysis result.
In the embodiment of the present application, the user interest point analysis network to be debugged represents a user interest point analysis network to be debugged, and the network compositions of the user interest point analysis network to be debugged and the debugged user interest point analysis network may be the same or different. The user interest point analysis network to be debugged may extend the network parameters of the debugged user interest point analysis network, in other words, the basic network parameters of the user interest point analysis network to be debugged may be the same as the network parameters of the debugged user interest point analysis network. In other embodiments, the underlying network parameters of the user point of interest analysis network to be commissioned are based on arbitrary settings.
As an implementation manner, the interest point analysis result may be an interest point mark indicating the interest point corresponding to the debug template, for example, the determined interest point and the newly introduced interest point include M interest points, and the interest point analysis result may be an array covering the interest point marks. Alternatively, the point of interest analysis results may indicate the probability (which may be represented by a probability array) that the debug template corresponds to the point of interest. The identified error value represents an error between the point of interest analysis result and the point of interest annotation information.
The user interest point analysis system loads each debugging template into a user interest point analysis network to be debugged respectively, the user interest point analysis network mines each debugging template to obtain second template characterization vectors corresponding to each debugging template respectively, then the user interest point analysis network performs interest point analysis processing on each debugging template according to the second template characterization vectors corresponding to each debugging template respectively to obtain interest point analysis results corresponding to each debugging template, and error determination (e.g. calculation based on cross entropy functions) is performed according to the interest point analysis results corresponding to each debugging template and interest point annotation information corresponding to each debugging template respectively to obtain identification error values.
S400: performing error determination according to a first template characterization vector and a second template characterization vector of the same debugging template to obtain a first target error value, and performing error determination according to second template characterization vectors respectively corresponding to the debugging template groups with different interest points to obtain a second target error value.
In this embodiment of the present application, the first target error value is associated with an error degree or a commonality coefficient (a value representing a similarity measurement result between a first template representation vector and a second template representation vector of the same debug template, so as to perform Knowledge Distillation, and the representation vector expected to be debugged for the user interest point analysis network mining is similar to the representation vector found by the debugged user interest point analysis network, so as to preserve knowledge acquired by the debugged user interest point analysis network.
Because debugging is performed according to the first target error value alone may reduce universality or generalization capability between newly introduced interest point types and past interest point types, a second target error value needs to be determined, the second target error value is associated with error degrees or commonality coefficients between second template characterization vectors respectively corresponding to debugging template groups with different interest points, and the second target error value is used for performing contrast learning so as to ensure that the mined characterization vectors have distinction from other characterization vectors.
In this embodiment of the present invention, for each debug template, a first template token vector and a second template token vector are obtained respectively, and then for each debug template loaded thereto, the user interest point analysis system may perform error determination according to the first template token vector and the second template token vector corresponding to the debug template, obtain sub error values corresponding to the debug template, and then obtain the first target error values based on the sub error values corresponding to each debug template respectively.
In addition, the user interest point analysis system performs error determination according to the second template characterization vectors corresponding to each two debugging templates (namely the debugging template group) with different interest points, so as to obtain sub-error values corresponding to each two debugging templates with different interest points, and then the second target error values are obtained based on the sub-error values.
For example, if the debug template set includes 3 debug templates, debug templates eg_a, debug templates eg_b, and debug templates eg_c, respectively. The debugging template Eg_A and the debugging template Eg_B have the same determined interest points, the debugging template Eg_C has new introduced interest points, the user interest point analysis system calculates the first template characterization vectors which are respectively mined by the 3 templates based on the debugged user interest point analysis network to be Vector-A, vector-B and Vector-C, the user interest point analysis system determines the sub-error value Loss-C of the debugging template Eg_C based on the Vector-A, vector-B and Vector-C, and determines the sub-error value Loss-C of the debugging template Eg_B based on the Vector-A, loss-B and the Vector-C 'based on the Vector-A and the Vector-A', and obtains the sub-error value Loss-A of the debugging template Eg_A based on the Vector-A and the Vector-A 'and the Vector-B' based on the Vector-B and the Vector-B 'and determines the sub-error value Loss-B of the debugging template Eg_B based on the Vector-C and the Vector-C' and the third sub-error value Loss-A 'and the third sub-error value Loss-B' based on the Vector-A and the Vector-B.
In one embodiment, the user interest point analysis system may determine, according to annotation information (e.g., a tag) of each debug template, a debug template group having different interest points in the process of obtaining the second target error value, and for each debug template group, the user interest point analysis system may perform error determination according to a second template token vector corresponding to the debug template group, to obtain a sub-error value corresponding to the debug template group.
As an implementation manner, since the debugged user interest point analysis network only learns the determined interest point, when the first error value is obtained, the user interest point analysis system may only obtain the debug template corresponding to the determined interest point. It should be noted that, the user interest point analysis network to be debugged learns the newly introduced interest point based on continuous learning, and the determined interest points in the debug template set are fewer, but the first target error value expects that the characterization vector of the network mining after updating and the characterization vector of the network mining before updating are similar, so that the user interest point analysis system can acquire the debug templates corresponding to the determined interest points and the debug templates corresponding to the newly introduced interest points in the process of acquiring the first target error value.
S500: and adjusting network parameter values of the interest point analysis network of the user to be debugged according to the first target error value, the second target error value and the identification error value, performing iterative debugging, and obtaining the interest point analysis network of the target user if the debugging cut-off requirement is met, wherein the interest point analysis network of the target user is configured to analyze and identify the determined interest point and the newly introduced interest point.
In the embodiment of the application, the user interest point analysis system can sum up the first target error value, the second target error value and the identification error value, complete statistics to obtain a sum-up error value, backward transfer the sum-up error value, adjust the network parameter of the user interest point analysis network to be debugged, then determine the user interest point analysis network obtained by the adjustment as the user interest point analysis network to be debugged, and repeatedly perform the above debugging process to perform iterative debugging until meeting the debugging cut-off requirement, thereby obtaining the debugged target user interest point analysis network. In the above process, the generalized manner is, for example, mean calculation, weighting calculation, or median calculation.
The network parameter of the user interest point analysis network can be adjusted based on gradient descent to complete optimization debugging of the network, and the debugging cut-off requirement is, for example, network convergence, for example, the network parameter is not changed obviously any more, the induction error value is minimum, or the debugging times meet the preset times. The target user interest point analysis network obtained by debugging learns a debugging template of the determined interest point and a debugging template of the newly introduced interest point during debugging, and can analyze and identify the determined interest point and the newly introduced interest point.
In the user interest point identification method based on big data analysis, when a network is debugged, a debug template set is obtained, the debug template set comprises debug templates with determined interest points corresponding to a debugged user interest point analysis network and debug templates with newly introduced interest points, first template characterization vectors corresponding to each debug template in the debug template set are respectively mined based on the debugged user interest point analysis network, second template characterization vectors corresponding to each debug template are respectively mined based on the to-be-debugged user interest point analysis network, interest point analysis is carried out according to each second template characterization vector, an identification error value is determined according to the obtained first interest point analysis result, error determination is carried out according to the first template characterization vectors and the second template characterization vectors of the same debug template, a first target error value is obtained, error determination is carried out according to the second template characterization vectors corresponding to the debug template groups with different interest points, network parameters of the user interest point analysis network are adjusted according to the first target error value, the second target error value and the identification error value, and the network interest point analysis is conducted according to the iteration interest point analysis requirements of the network, and if the interest point analysis of the network is improved, and the interest point analysis interest point of the network is improved is achieved, and if the interest point analysis of the network is required to be debugged is improved.
As an implementation manner, performing error determination according to second template characterization vectors respectively corresponding to debug template groups with different interest points to obtain a second target error value, including: respectively constructing a debugging template group by each debugging template and each debugging template in the debugging template set to obtain a plurality of debugging template groups; determining a common coefficient according to the second template characterization vectors respectively corresponding to the debugging template groups, and acquiring sub-error values respectively corresponding to the debugging template groups according to the determined common coefficient; and inducing sub-error values corresponding to the target debugging template group to obtain a second target error value, wherein the target debugging template group is a debugging template group of debugging templates with different interest points.
For example, the user interest point analysis system traverses all the debug templates of the debug template set, constructs a debug template group from each of the debug templates and each of the debug templates of the debug template set, thus obtaining a plurality of debug template groups, determines a common coefficient according to respective second debug template characterization vectors of two debug templates in the debug template group for each debug template group, and determines a sub-error value corresponding to the debug template group according to the determined common coefficient. And the sub-error value corresponding to the debugging template group is positively correlated with the determined common coefficient. Because the purpose of the second target error value is to perform contrast learning, after determining the sub-error values corresponding to the debug template groups respectively, the user interest point analysis system only generalizes the sub-error values of the target template pairs of the debug templates with different interest points, and determines the generalized error value obtained by generalization as the second target error value.
Specifically, for each debug template group, the user interest point analysis system can determine a commonality coefficient according to the respective second debug template characterization vectors of the two debug templates in the debug template group, so as to obtain the commonality coefficient corresponding to the debug template group. The calculation of the commonality coefficient may be based on determining a distance between two token vectors, e.g. a cosine distance, a euclidean distance, etc. And acquiring a sub-error value based on the commonality coefficient, and then inducing a second target error value corresponding to the target template group, wherein the acquired second target error value can better embody the discrimination capability of the debugging templates of different interest points.
As an implementation manner, obtaining the sub-error values corresponding to each debug template group according to the determined commonality coefficient may specifically include: for the common coefficient corresponding to each debugging template, the common coefficient is subjected to difference with a preset common coefficient to obtain a target coefficient difference; when the common coefficient is smaller than the preset common coefficient, determining the minimum error value as a sub-error value corresponding to the debugging template; and when the common coefficient is not smaller than the preset common coefficient, determining the target coefficient difference as a sub-error value corresponding to the debugging template. For example, a preset common coefficient is determined in advance, the preset common coefficient is determined to be the maximum common coefficient among the different interest point debugging templates, and the network parameter of the interest point analysis network of the user to be debugged is adjusted to the direction that the common coefficient among the different interest point debugging templates is smaller than the maximum common coefficient. And for the common coefficient corresponding to each debugging template group, the user interest point analysis system performs difference between the common coefficient and a preset common coefficient to obtain a target coefficient difference, when the common coefficient is smaller than the preset common coefficient, the minimum error value is determined as a sub-error value corresponding to the debugging template, and when the common coefficient is not smaller than the preset common coefficient, the target coefficient difference is determined as the sub-error value corresponding to the debugging template. In the debugging process, the network parameter of the interest point analysis network of the user to be debugged is adjusted in the direction of error reduction, in other words, the direction of the sub-error value corresponding to the debugging template determined by the minimum error value is adjusted, so that the common coefficient among the debugging templates of different interest points is smaller than the preset common coefficient, and the excavated characterization vectors and the rest characterization vectors have higher discrimination capability.
Based on the above process, when determining the sub-error value, if the common coefficient is smaller than the preset common coefficient, determining the minimum error value as the sub-error value corresponding to the debugging template, and if the common coefficient is not smaller than the preset common coefficient, determining the target coefficient difference as the sub-error value corresponding to the debugging template, so that the discrimination capability of the target user interest point analysis network obtained by debugging on the characterization vectors mined by different interest point templates can be ensured, and the generalization capability of the network is enhanced.
As an implementation manner, the step of summarizing the sub-error values corresponding to the target debug template group to obtain the second target error value may specifically include: aiming at each debugging template group, acquiring interest point annotation information corresponding to the debugging templates included in the debugging template group, and acquiring an illustrative result corresponding to the debugging template group according to the interest point annotation information; the method comprises the steps that an illustrative result is determined based on an illustrative function (indicator function), when interest point annotation information corresponding to a debugging template included in a debugging template group is the same, the illustrative result determined based on the illustrative function is a first character, and when interest point annotation information corresponding to the debugging template included in the debugging template group is the same, the illustrative result determined based on the illustrative function is a second character; when the obtained illustrative result is determined to be a first character, carrying out a retention operation on the sub-error value corresponding to the debugging template group to obtain a target sub-error value corresponding to the debugging template group, and when the obtained illustrative result is determined to be a second character, carrying out a cleaning operation on the sub-error value corresponding to the debugging template group to obtain a target sub-error value corresponding to the debugging template group; and inducing target sub-error values corresponding to the debugging template groups respectively to obtain second target error values. The interest point annotation information is used for indicating the interest point type of the debugging template and is used for monitoring in the debugging process.
In the above process, the exemplary result output by the exemplary function is a first character or a second character, where the first character and the second character correspond to different values (e.g. 0 or 1), and then the consistency of the interest point annotation information corresponding to the debug template covered by the debug template group can be determined by outputting the exemplary result, when the obtained exemplary result is the first character, the interest point annotation information corresponding to the debug template included in the debug template group is the same, and then the user interest point analysis system performs a retention operation on the sub-error value corresponding to the debug template group, so that the sub-error value is used as a component of the second target error value, and when the obtained exemplary result is the second character, the interest point annotation information corresponding to the debug template included in the debug template group is different, and then the user interest point analysis system performs a cleaning operation (e.g. shielding) on the sub-error value corresponding to the debug template group, so as to avoid the disturbance of the sub-error value on the second target error value.
Specifically, the user interest point analysis system takes a product of a sub-error value corresponding to the debug template group and an illustrative result corresponding to the debug template, when the illustrative result is a first character, the sub-error value is taken as a target sub-error value of the debug template based on the product taken by the first character, when the illustrative result is a second character, the target sub-error value of the debug template is 0 based on the product taken by the second character, and finally the generalized second target error value only comprises the sub-error value corresponding to the target debug template group, wherein the target debug template group is the debug template group of the debug templates with different interest points.
Based on the above process, the sub-error values of the debug template group of the debug templates with the same interest point type can be cleaned by obtaining the illustrative result, so that the obtained second target error value is more accurate, and the recognition capability of the target user interest point analysis network is enhanced.
As an implementation manner, the embodiment of the present application further provides a method for debugging a user interest point analysis network, which may specifically include:
(1) A set of debug templates is obtained.
The debug template set comprises debug templates with determined points of interest corresponding to the debugged user point of interest analysis network and debug templates with newly introduced points of interest.
(2) And respectively mining first template characterization vectors corresponding to each debugging template in the debugging template set based on the debugged user interest point analysis network.
(3) And respectively mining second template characterization vectors corresponding to each debugging template based on the interest point analysis network of the user to be debugged, carrying out interest point analysis according to each second template characterization vector, and determining an identification error value according to each obtained first interest point analysis result.
(4) Performing error determination according to a first template characterization vector and a second template characterization vector of the same debugging template to obtain a first target error value, and performing error determination according to second template characterization vectors respectively corresponding to the debugging template groups with different interest points to obtain a second target error value.
(5) And carrying out error determination according to the first template characterization vector and the second template characterization vector corresponding to the debugging template groups with different interest points to obtain a third target error value.
The user interest point analysis system performs error determination according to the first template characterization vector and the second template characterization vector corresponding to each two debugging templates with different interest points, obtains sub-error values corresponding to each two debugging templates with different interest points, and then generalizes the sub-error values to obtain a third target error value. For example, the debug template group includes two debug templates, namely a debug template eg_a and a debug template eg_b, and as an implementation manner, the first template characterization vector and the second template characterization vector corresponding to the debug template group may be the first template characterization vector and the second template characterization vector corresponding to the debug template eg_a and the second template characterization vector corresponding to the debug template eg_b, and the user interest point analysis system may determine an error according to the first template characterization vector and the second template characterization vector corresponding to the debug template eg_a to obtain a sub-error value of the debug template group; as still another embodiment, the first template token vector and the second template token vector corresponding to the debug template group may be a second template token vector corresponding to the debug template eg_a and a first template token vector corresponding to the debug template eg_b, and the user interest point analysis system may perform error determination according to the second template token vector corresponding to the debug template eg_a and the first template token vector corresponding to the debug template eg_b to obtain a sub error value of the debug template group; or, the user interest point analysis system may determine that the first template token vector and the second template token vector corresponding to the debug template group may include the first template token vector and the second template token vector corresponding to the debug template eg_a and the second template token vector corresponding to the debug template eg_b, and the second template token vector and the first template token vector corresponding to the debug template eg_a and the second template token vector corresponding to the debug template eg_b, and then determine an error value according to the first template token vector and the second template token vector corresponding to the debug template eg_a and the second template token vector corresponding to the debug template eg_b, and determine that an error value is obtained according to the error determination of the second template token vector and the first template token vector corresponding to the debug template eg_b, and finally sum up the two error values to obtain the error value of the debug template group.
In one embodiment, in the error determination process, for each group of a template characterization vector and a second template characterization vector for error determination, a common coefficient of the two template characterization vectors is determined, and a corresponding sub-error value is determined according to the determined common coefficient. As an implementation manner, for each obtained common coefficient, the common coefficient is subjected to difference with a preset common coefficient to obtain a target coefficient difference; when the common coefficient is smaller than a preset common coefficient, determining the minimum error value as a corresponding sub-error value; and when the common coefficient is not smaller than the preset common coefficient, determining the target coefficient difference as a corresponding sub-error value.
(6) The first target error value, the second target error value, the third target error value and the identification error value are summed up, the summed-up error value is obtained, the network parameter of the user interest point analysis network to be debugged is adjusted according to the summed-up error value, and iterative debugging is carried out.
Specifically, in the induction process, the user interest point analysis system can induce the second target error value and the third target error value based on the following formula:
Loss＝∑F(x)(Max((V 1 ,V 2 )-β，0)+Max((V 1 ,V 2 )-β，0))
wherein Loss is an error value obtained by inducing the second target error value and the third target error value, and F (x) is an indication function; (V) 1 ,V 2 ) For the first characterization vector and the second characterization vector corresponding to the debugging template group, beta is a preset commonality coefficient; and if the annotation information of the debug template groups is the same, the illustrative result output by the illustrative function is equal to 0, and if the annotation information of the debug template groups is different, the illustrative result output by the illustrative function is equal to 1.
Because the first template characterization vector is a characterization vector obtained by extracting the network before updating, and the second template characterization vector is a characterization vector obtained by extracting the network after updating, the adoption of the third target error value can ensure that the updated network and the network before updating have the identification capability on the characterization vectors mined by different interest points, so that the updated network better identifies the newly introduced interest points and the determined interest points, and the generalization capability of the network is enhanced. As an implementation manner, performing error determination according to a first template characterization vector and a second template characterization vector of the same debug template to obtain a first target error value, including: determining a common coefficient according to a first template characterization vector and a second template characterization vector which correspond to each debugging template respectively, and acquiring respective sub-error values of each debugging template according to the determined common coefficient; the sub-error value of the debugging template is reversely related to the commonality coefficient corresponding to the debugging template; and summarizing the respective sub-error values of each debugging template to obtain a first target error value.
For example, the first target error value is determined based on the following formula:
Loss1＝∑1-(V 1 ,V 2 )
loss1 is a first target error value, V1 is a second template characterization vector of the point of interest analysis network mining of the user to be debugged, V 2 Analyzing a network mined first template token vector for a debugged user interest point, (V) 1 ,V 2 ) And (3) representing the common coefficient between the vector for the first template and the vector for the second template.
As one embodiment, the determined points of interest include a plurality of points of interest, and the debug template set is constructed by: acquiring a basic template set in a preset data space (such as a data space opened in a memory), wherein the basic template set comprises reference debugging templates corresponding to the determined interest points, and the scattering condition of the reference debugging templates corresponding to the determined interest points in the characterization vector value domain corresponding to the determined interest points is associated with the scattering condition of the basic template set corresponding to the determined interest points in the characterization vector value domain; and acquiring a new introduced template set corresponding to the new introduced interest point, and constructing a debugging template set according to the new introduced template set and the basic template set.
The basic template set corresponding to the determined interest point represents a debugging template set with the determined interest point, which is adopted when the debugged user interest point analysis network is debugged, and the representation vector value field corresponding to the determined interest point represents the representation vector value field formed by the template representation vectors corresponding to the templates in the basic template set corresponding to the determined interest point. The dispersion of the reference debug templates corresponding to the determined points of interest in the token vector value domain corresponding to the determined points of interest and the dispersion of the set of base templates corresponding to the determined points of interest in the token vector value domain are associated (matched to each other), in other words, the distribution of the reference debug templates corresponding to the determined points of interest in the token vector value domain is the position of the template token vector of the reference debug templates in the token vector value domain near the center of each position in the token vector value domain. In the embodiment of the application, the preset data space can store some reference debugging templates for each determined interest point, the reference debugging templates are selected based on the value domain structure of the representation vector value domain corresponding to the determined interest point, high-quality data in the basic template set is reserved, the reference debugging templates are conveniently acquired in the preset data space during continuous learning, the newly introduced template set corresponding to the newly introduced interest point is acquired, and the newly introduced template set is combined with the reference debugging templates in the basic template set to obtain the debugging interest point set for continuous learning. Because the reference debugging templates of the determined interest points are stored in the preset data space, the scattering condition of the reference debugging templates in the representation vector value domain corresponding to the determined interest points is associated with the scattering condition of the basic template set corresponding to the determined interest points in the representation vector value domain, for each determined interest point, the templates rich in information can be obtained, the network can be debugged based on the templates, the network can learn the template representation vectors of the determined interest points, and the analysis capability of the network on the determined interest points is maintained.
As an implementation manner, before the basic template set is acquired in the preset data space, the method provided in the embodiment of the present application further includes: acquiring a basic template set corresponding to each determined interest point respectively; the basic template set comprises an analyzed debugging template corresponding to the debugged user interest point analysis network; grouping template characterization vectors corresponding to the analyzed and debugged templates in the basic template set (for example, clustering based on a clustering algorithm) to obtain a plurality of grouping results corresponding to the determined interest points; screening template characterization vectors, of which the vector distances between the template characterization vectors and the respective grouping centroids meet the preset distance requirement (for example, the vector distances between the template characterization vectors and the grouping centroids of the grouping result are smaller than the preset distance), in each grouping result, and taking the analyzed debugging templates corresponding to the screened template characterization vectors as reference debugging templates corresponding to the determined interest points. The analyzed debugging templates corresponding to the debugged user interest point analysis network represent the debugging templates adopted during debugging of the debugging network, and the basic template set corresponding to the determined interest points is the debugging template with the determined interest points adopted during debugging to obtain the debugged user interest point analysis network.
For each basic template set corresponding to the determined interest point, the user interest point analysis system can extract the template characterization vector of each analyzed debugging template in the basic template set based on the debugged user interest point analysis network, group the template characterization vectors, obtain a plurality of grouping results after grouping the basic template set corresponding to each determined interest point, screen at least one template characterization vector with the distance between the grouping results and the respective grouping centroid meeting the preset distance requirement, and the screened template characterization vector represents the corresponding analyzed debugging template, so that the analyzed debugging template corresponding to the screened template characterization vector can be determined as the reference debugging template with the determined interest point of the analyzed debugging template. Through the process, based on grouping, template characterization vectors with the vector distances of the centroid of each grouping meeting the preset distance requirement are respectively screened in each grouping result, an analyzed debugging template corresponding to the screened template characterization vectors is determined to be a reference debugging template corresponding to the determined interest point, and the screened templates belong to different interest points after grouping, have abundant information and have the property of retaining the existing templates.
As an implementation manner, grouping template characterization vectors corresponding to each analyzed and debugged template in the basic template set to obtain a plurality of grouping results corresponding to the determined interest points, which specifically includes: determining a first interest point statistics value (namely a corresponding number) corresponding to the determined interest points, and determining a target interest point statistics value during grouping according to the first interest point statistics value and the data space size of a preset data space; grouping template characterization vectors respectively corresponding to the analyzed and debugged templates in the basic template set according to the target interest point statistics values to obtain a plurality of grouping results corresponding to the determined interest points and matched with the target interest point statistics values. The data space size of the preset data space represents the maximum number of templates that the preset data space can store.
Specifically, assuming that the data space size of the preset data space is Q, the statistics of the first points of interest corresponding to the determined points of interest is P, and for each determined point of interest, the statistics of the target points of interest in grouping is Q/P. After the statistics of the target points of interest is obtained, aiming at each basic template set, template characterization vectors respectively corresponding to the analyzed and debugged templates of the basic template set are grouped according to the statistics of the target points of interest, and the number of the obtained grouping results is the same as that of the statistics of the target points of interest. Based on the first interest point statistics value corresponding to the determined interest point, determining a target interest point statistics value in grouping according to the first interest point statistics value and the data space size of the preset data space, grouping template characterization vectors respectively corresponding to each analyzed debugging template in the basic template set according to the target interest point statistics value, so that the reference debugging templates of each determined interest point can be uniformly and maximally stored in the preset data space.
As an implementation manner, the newly introduced interest point includes a plurality of newly introduced interest points, and after obtaining the interest point analysis network of the target user, the method provided by the embodiment of the application further includes: determining a second interest point statistics corresponding to the newly introduced interest point, and cleaning a plurality of reference debugging templates in the basic template set in a preset data space according to the second interest point statistics; determining a reference debugging template corresponding to the new introduced interest point in a new introduced template set corresponding to the new introduced interest point; and saving the reference debugging template corresponding to the newly introduced interest point to a preset data space.
The new introduced template set corresponding to the new introduced interest point represents a debugging template with the new introduced interest point adopted when the target user interest point analysis network is obtained through debugging, and the scattering condition of the reference debugging template of the new introduced interest point in the characterization vector value domain corresponding to the new introduced interest point is associated (e.g. matched) with the scattering condition of the new introduced template set corresponding to the new introduced interest point in the characterization vector value domain, which has been described previously and will not be repeated here.
And acquiring a target user interest point analysis network, and storing template data of the newly introduced interest points adopted in the debugging process into a preset data space, so that further continuous learning is facilitated. However, because the data space size of the preset data space is fixed, some data is cleaned off, so that space is vacated, template data of the newly introduced interest point is stored, the user interest point analysis system can determine a second interest point statistics of the newly introduced interest point, a plurality of reference debugging templates in the basic template set are cleaned off in the preset data space according to the second interest point statistics, and the more the second interest point statistics are, the more templates are to be cleaned off. For each new introduced interest point, the user interest point analysis system determines a reference debugging template in a new introduced template set corresponding to the new introduced interest point, and then stores the reference debugging templates corresponding to the new introduced interest points into a preset data space to finish updating the preset data space. Based on the method, a plurality of reference debugging templates in a basic template set are cleaned in a preset data space according to the second interest point statistics, the reference debugging templates corresponding to the new introduction interest points are determined in the new introduction template set corresponding to the new introduction interest points, the reference debugging templates corresponding to the new introduction interest points are stored in the preset data space, and templates required for continuous learning can be stored based on the preset data space with a certain data space size.
As an implementation manner, cleaning the plurality of reference debugging templates in the basic template set in the preset data space according to the second interest point statistics value may specifically include: determining the target template quantity of the templates to be cleaned corresponding to each determined interest point according to the second interest point statistics; for each determined interest point, determining a representative characterization vector of a characterization vector value range corresponding to the determined interest point according to an analyzed debugging template corresponding to the determined interest point; respectively obtaining vector distances between template characterization vectors and representative characterization vectors respectively corresponding to each reference debugging template; and cleaning the reference template characterization vectors matched with the number of the target templates from the reference template characterization vectors corresponding to the determined interest points according to the vector distances respectively corresponding to the reference debugging templates.
For example, the user interest point analysis system determines the number of target templates of the templates to be cleaned corresponding to each determined interest point according to the second interest point statistics, determines a template characterization vector mean value according to the template characterization vector of the analyzed debugging template corresponding to each determined interest point, determines the template characterization vector mean value as a representative characterization vector of the characterization vector value field corresponding to the determined interest point, respectively acquires vector distances between the template characterization vector and the representative characterization vector corresponding to each reference debugging template corresponding to the determined interest point, respectively arranges the reference debugging templates corresponding to the determined interest point according to the increasing times according to the vector distances corresponding to each reference debugging template, and cleans the reference debugging templates in front according to the arrangement result, wherein the number of the cleaned reference debugging templates is the same as the number of the target templates. The template characterization vectors of all the analyzed debugging templates corresponding to the determined interest points can be subjected to mean value calculation to obtain a template characterization vector mean value. Based on the method, the representative characterization vectors of the characterization vector value fields corresponding to the determined interest points are determined, the vector distances between the template characterization vectors and the representative characterization vectors corresponding to the reference debugging templates are obtained respectively, the templates are cleaned according to the vector distances corresponding to the reference debugging templates, and the stored templates are similar to the representative characterization vectors so as to maximize the data scattering property of the remaining characterization vector value fields.
As one implementation, the first interest point analysis result includes a first inference confidence level corresponding to the determined interest point; before adjusting the network parameter of the user interest point analysis network to be debugged according to the first target error value, the second target error value and the identification error value and performing iterative debugging, the method provided by the embodiment of the application further comprises the following steps: based on the debugged user interest point analysis network, interest point analysis is carried out according to the first template characterization vector corresponding to each debugging template respectively, and a second interest point analysis result of each debugging template is obtained; the second interest point analysis result comprises a second reasoning confidence degree corresponding to the determined interest point; for each debugging template, carrying out standardization operation on the first reasoning confidence coefficient and the second reasoning confidence coefficient of the debugging template under the same refined variable to obtain a first target confidence coefficient corresponding to the first reasoning confidence coefficient and a second target confidence coefficient corresponding to the second reasoning confidence coefficient; determining a refining error value according to the first target confidence coefficient and the second target confidence coefficient to obtain a refining error value of the debugging template; summarizing the extraction error values of each debugging template to obtain a target extraction error value; adjusting the network parameter of the user interest point analysis network to be debugged according to the first target error value, the second target error value and the identification error value, and performing iterative debugging, including: inducing a first target error value, a second target error value, a target extraction error value and an identification error value to obtain an induced error value; and adjusting network parameters of the interest point analysis network of the user to be debugged according to the induced error value and performing iterative debugging. The method for obtaining the target extraction error value is, for example, calculated based on a general knowledge distillation error determination formula, which is not limited herein, and the above process is implemented by knowledge distillation, where the extraction error value is a distillation error, and the extraction variable may be a temperature value in knowledge distillation. After the target extraction error value is obtained, the first target error value, the second target error value, the target extraction error value and the identification error value can be generalized to obtain a generalized error value, and then the network parameter of the interest point analysis network of the user to be debugged is adjusted according to the generalized error value. Based on the method, the target extraction error value is obtained, so that the interest point analysis network of the user to be debugged can be promoted to maintain the probability dispersion property of the debugged network during learning, and the universality and generalization capability of the network are enhanced.
Referring to fig. 4, a functional module architecture diagram of a user interest point identifying apparatus 110 according to an embodiment of the present invention is provided, where the user interest point identifying apparatus 110 may be used to perform a user interest point identifying method based on big data analysis, and the user interest point identifying apparatus 110 includes:
the data acquisition module 111 is configured to acquire user exhibition behavior data in a preset behavior data storage space in response to a user analysis instruction;
the network calling module 112 is configured to load the user exhibition behavior data into a user interest point analysis network, and perform interest point analysis on the user exhibition behavior data based on the user interest point analysis network;
the result obtaining module 113 is configured to obtain a point of interest analysis result corresponding to the user exhibition behavior data output by the user point of interest analysis network;
the network debugging module 114 is used for debugging the user interest point analysis network. The user interest point analysis network is obtained by adjusting network parameters of a user interest point analysis network to be debugged according to a first target error value, a second target error value and an identification error value, wherein the first target error value is obtained by carrying out error determination according to a first template characterization vector and a second template characterization vector of the same debugging template, the second target error value is obtained by carrying out error determination according to second template characterization vectors respectively corresponding to debugging template groups with different interest points, the first template characterization vector of the debugging template is obtained based on extraction of the debugged user interest point analysis network, the second template characterization vector of the debugging template is obtained based on extraction of the user interest point analysis network to be debugged, the identification error value is obtained according to interest point analysis results generated by carrying out interest point analysis on the second template characterization vector of the debugging template according to the user interest point analysis network to be debugged, and the debugging template set comprises a debugging template with determined interest points corresponding to the debugged user interest point analysis network and a debugging template with new introduced interest points.
Since in the above embodiment, the detailed description has been made of the method for identifying the user interest point based on big data analysis provided in the embodiment of the present invention, and the principle of the device for identifying the user interest point 110 is the same as that of the method, the execution principle of each module of the device for identifying the user interest point 110 will not be described in detail here.
In the several embodiments provided in this application, it should be understood that the disclosed apparatus and method may be implemented in other manners as well. The apparatus embodiments described above are merely illustrative, for example, of the flowcharts and block diagrams in the figures that illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
In addition, functional modules in the embodiments of the present invention may be integrated together to form a single part, or each module may exist alone, or two or more modules may be integrated to form a single part.
The functions, if implemented in the form of software functional modules and sold or used as a stand-alone product, may be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, an internet of things data server, or a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
It is noted that relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
It is to be understood that the terminology which does not make a noun interpretation with respect to the above description is not to be interpreted as a noun interpretation, and that the skilled person can unambiguously ascertain the meaning to which it refers from the above disclosure. The foregoing of the disclosure of the embodiments of the present application will be apparent to and complete with respect to those skilled in the art. It should be appreciated that the process of deriving and analyzing technical terms not explained based on the above disclosure by those skilled in the art is based on what is described in the present application, and thus the above is not an inventive judgment of the overall scheme.
While the basic concepts have been described above, it will be apparent to those skilled in the art that the foregoing detailed disclosure is by way of example only and is not intended to be limiting. Although not explicitly described herein, various modifications, improvements, and adaptations may occur to one skilled in the art. Such modifications, improvements, and modifications are intended to be suggested within this application, and are therefore within the spirit and scope of the exemplary embodiments of this application.
It should also be appreciated that in the foregoing description of the embodiments of the present application, various features are sometimes grouped together in a single embodiment, figure, or description thereof for the purpose of streamlining the disclosure aiding in the understanding of at least one of the embodiments of the invention. This method of disclosure, however, is not intended to imply that more features than are presented in the claims are required for the subject application. Indeed, less than all of the features of a single embodiment disclosed above.
Claims (10)
1. A method for identifying points of interest of a user based on big data analysis, characterized in that it is applied to a system for identifying points of interest of a user, said method comprising:
responding to a user analysis instruction, and acquiring user exhibition behavior data in a preset behavior data storage space;
loading the user exhibition behavior data into a user interest point analysis network, and carrying out interest point analysis on the user exhibition behavior data based on the user interest point analysis network;
the user interest point analysis network is obtained by adjusting network parameters of a user interest point analysis network to be debugged according to a first target error value, a second target error value and an identification error value, wherein the first target error value is obtained by carrying out error determination according to a first template characterization vector and a second template characterization vector of the same debugging template, the second target error value is obtained by carrying out error determination according to second template characterization vectors respectively corresponding to debugging template groups with different interest points, the first template characterization vector of the debugging template is obtained based on extraction of the debugged user interest point analysis network, the second template characterization vector of the debugging template is obtained based on extraction of the user interest point analysis network to be debugged, the identification error value is obtained according to interest point analysis results generated by carrying out interest point analysis on the second template characterization vector of the debugging template by the user interest point analysis network to be debugged, and the debugging template set comprises a debugging template with determined interest points corresponding to the debugged user interest point analysis network and a debugging template with new introduction points;
And acquiring an interest point analysis result corresponding to the user exhibition behavior data output by the user interest point analysis network.
2. The method for identifying points of interest of a user based on big data analysis of claim 1, further comprising a step of debugging the network for analyzing points of interest of the user, comprising:
acquiring a debugging template set, wherein the debugging template set comprises a debugging template with determined interest points corresponding to a debugged user interest point analysis network and a debugging template with newly introduced interest points;
respectively mining first template characterization vectors corresponding to each debugging template in the debugging template set based on the debugged user interest point analysis network;
respectively mining second template characterization vectors corresponding to the debugging templates based on a user interest point analysis network to be debugged, carrying out interest point analysis according to the second template characterization vectors, and determining identification error values according to the obtained first interest point analysis results;
performing error determination according to a first template characterization vector and a second template characterization vector of the same debugging template to obtain a first target error value, and performing error determination according to second template characterization vectors respectively corresponding to the debugging template groups with different interest points to obtain a second target error value;
And adjusting network parameter values of the interest point analysis network of the user to be debugged according to the first target error value, the second target error value and the identification error value, performing iterative debugging, and obtaining a target user interest point analysis network if the debugging cut-off requirement is met, wherein the target user interest point analysis network is configured to analyze and identify the determined interest point and the newly introduced interest point.
3. The method of claim 2, wherein the performing error determination according to the second template token vectors respectively corresponding to the debug template groups having different points of interest to obtain a second target error value includes:
respectively constructing each debugging template and each debugging template in the debugging template set into a debugging template group to obtain a plurality of debugging template groups;
determining a common coefficient according to the second template characterization vectors respectively corresponding to the debugging template groups, and acquiring sub-error values respectively corresponding to the debugging template groups according to the determined common coefficient;
and inducing sub-error values corresponding to the target debugging template group to obtain a second target error value, wherein the target debugging template group is a debugging template group of the debugging templates with different interest points.
4. A method according to claim 3, wherein the obtaining the sub-error values respectively corresponding to the debug template groups according to the determined commonality coefficients includes:
for the common coefficient corresponding to each debugging template, the common coefficient is subjected to difference with a preset common coefficient to obtain a target coefficient difference;
when the common coefficient is smaller than the preset common coefficient, determining a minimum error value as a sub-error value corresponding to the debugging template;
when the common coefficient is not smaller than the preset common coefficient, determining the target coefficient difference as a sub-error value corresponding to the debugging template;
the inducing the sub-error value corresponding to the target debugging template group to obtain a second target error value comprises the following steps:
for each debugging template group, obtaining interest point annotation information corresponding to a debugging template included in the debugging template group, and calculating an illustrative result corresponding to the debugging template group according to the interest point annotation information; the method comprises the steps that an illustrative result is determined based on an illustrative function, when interest point annotation information corresponding to a debugging template included in a debugging template group is the same, the illustrative result determined based on the illustrative function is a first character, and when interest point annotation information corresponding to the debugging template included in the debugging template group is the same, the illustrative result determined based on the illustrative function is a second character;
When the obtained illustrative result is determined to be a first character, carrying out a retention operation on the sub-error value corresponding to the debugging template group to obtain a target sub-error value corresponding to the debugging template group, and when the obtained illustrative result is determined to be a second character, carrying out a cleaning operation on the sub-error value corresponding to the debugging template group to obtain a target sub-error value corresponding to the debugging template group;
and inducing target sub-error values corresponding to the debugging template groups respectively to obtain second target error values.
5. The method of any of claims 1-4, wherein prior to said adjusting the network parameters of the user interest point analysis network to be debugged and iteratively debugging in accordance with the first target error value, the second target error value, and the identified error value, the method further comprises:
performing error determination according to the first template characterization vector and the second template characterization vector corresponding to the debugging template groups with different interest points to obtain a third target error value;
the adjusting the network parameter of the user interest point analysis network to be debugged according to the first target error value, the second target error value and the identification error value and performing iterative debugging comprises the following steps:
Summarizing the first target error value, the second target error value, the third target error value and the identification error value to obtain a summarized error value;
and adjusting the network parameter values of the interest point analysis network of the user to be debugged according to the induced error value, and performing iterative debugging.
6. The method of claim 2, wherein the performing error determination based on the first template token vector and the second template token vector of the same debug template to obtain a first target error value comprises:
determining a common coefficient according to a first template characterization vector and a second template characterization vector which correspond to each debugging template respectively, and acquiring respective sub-error values of each debugging template according to the determined common coefficient; the sub-error value of the debugging template is reversely associated with the commonality coefficient corresponding to the debugging template;
summarizing respective sub-error values of each debugging template to obtain a first target error value;
the determined interest points comprise a plurality of interest points, and the debug template set is constructed by the following steps:
acquiring a basic template set in a preset data space; the basic template set comprises a reference debugging template corresponding to the determined interest point, and the scattering condition of the reference debugging template corresponding to the determined interest point in the characterization vector value domain corresponding to the determined interest point is associated with the scattering condition of the basic template set corresponding to the determined interest point in the characterization vector value domain;
Acquiring a new introduced template set corresponding to the new introduced interest point, and constructing a debugging template set according to the new introduced template set and the basic template set;
before the basic template set is acquired in the preset data space, the method further comprises:
acquiring a basic template set corresponding to each determined interest point respectively; the basic template set comprises an analyzed debugging template corresponding to the debugged user interest point analysis network;
grouping template characterization vectors respectively corresponding to all analyzed and debugged templates in the basic template set for each basic template set corresponding to the determined interest points to obtain a plurality of grouping results corresponding to the determined interest points;
screening template characterization vectors with vector distances meeting the preset distance requirement from the respective grouping results, and determining an analyzed debugging template corresponding to the screened template characterization vectors as a reference debugging template corresponding to the determined interest point.
7. The method of claim 6, wherein grouping template token vectors corresponding to each of the analyzed debug templates in the base set of templates to obtain a plurality of grouping results corresponding to the determined points of interest, comprises:
Determining a first interest point statistics value corresponding to the determined interest point, and determining a target interest point statistics value during grouping according to the first interest point statistics value and the data space size of the preset data space;
grouping template characterization vectors respectively corresponding to all analyzed and debugged templates in the basic template set according to the target interest point statistics values to obtain a plurality of grouping results corresponding to the determined interest points and matched with the target interest point statistics values.
8. The method of claim 6, wherein the newly introduced point of interest comprises a plurality, and wherein after obtaining the target user point of interest analysis network, the method further comprises:
determining a second interest point statistics corresponding to the newly introduced interest point, and cleaning a plurality of reference debugging templates in the basic template set in the preset data space according to the second interest point statistics;
determining a reference debugging template corresponding to the new introduced interest point in a new introduced template set corresponding to the new introduced interest point;
storing the reference debugging template corresponding to the newly introduced interest point into the preset data space;
The step of cleaning the plurality of reference debugging templates in the basic template set in the preset data space according to the second interest point statistics comprises the following steps:
determining the target template quantity of the templates to be cleaned corresponding to each determined interest point according to the second interest point statistics value;
for each determined interest point, determining a representative characterization vector of a characterization vector value range corresponding to the determined interest point according to an analyzed debugging template corresponding to the determined interest point;
respectively obtaining vector distances between template characterization vectors and the representative characterization vectors respectively corresponding to the reference debugging templates;
and cleaning the reference template characterization vectors matched with the target template number from the reference template characterization vectors corresponding to the determined interest points according to the vector distances respectively corresponding to the reference debugging templates.
9. The method of claim 8, wherein the first point of interest analysis result includes a first inference confidence level corresponding to the determined point of interest;
before the adjusting the network parameter of the user interest point analysis network to be debugged according to the first target error value, the second target error value and the identification error value and performing iterative debugging, the method further comprises:
Based on the debugged user interest point analysis network, performing interest point analysis according to the first template characterization vectors respectively corresponding to the debug templates to obtain respective second interest point analysis results of the debug templates; the second interest point analysis result comprises a second reasoning confidence degree corresponding to the determined interest point;
for each debugging template, carrying out standardization operation on the first reasoning confidence coefficient and the second reasoning confidence coefficient of the debugging template under the same refined variable to obtain a first target confidence coefficient corresponding to the first reasoning confidence coefficient and a second target confidence coefficient corresponding to the second reasoning confidence coefficient;
determining a refining error value according to the first target confidence coefficient and the second target confidence coefficient to obtain a refining error value of the debugging template;
summarizing the extraction error values of each debugging template to obtain a target extraction error value;
the adjusting the network parameter of the user interest point analysis network to be debugged according to the first target error value, the second target error value and the identification error value and performing iterative debugging comprises the following steps:
summarizing the first target error value, the second target error value, the target refinement error value and the identification error value to obtain a summarizing error value;
And adjusting the network parameter values of the interest point analysis network of the user to be debugged according to the induced error value, and performing iterative debugging.
10. A user point of interest analysis system comprising a processor and a memory, the memory storing a computer program for execution by the processor to implement the method of any of claims 1-9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211740018.6A CN116362782A (en) | 2022-12-31 | 2022-12-31 | User interest point identification method and system based on big data analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211740018.6A CN116362782A (en) | 2022-12-31 | 2022-12-31 | User interest point identification method and system based on big data analysis |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116362782A true CN116362782A (en) | 2023-06-30 |
Family
ID=86930472
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211740018.6A Pending CN116362782A (en) | 2022-12-31 | 2022-12-31 | User interest point identification method and system based on big data analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116362782A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116975455A (en) * | 2023-09-24 | 2023-10-31 | 太仓市律点信息技术有限公司 | User interest identification method and device applied to artificial intelligence |
-
2022
- 2022-12-31 CN CN202211740018.6A patent/CN116362782A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116975455A (en) * | 2023-09-24 | 2023-10-31 | 太仓市律点信息技术有限公司 | User interest identification method and device applied to artificial intelligence |
| CN116975455B (en) * | 2023-09-24 | 2023-12-22 | 太仓市律点信息技术有限公司 | User interest recognition method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110363220B (en) | Behavior class detection method and device, electronic equipment and computer readable medium | |
| CN110807491A (en) | License plate image definition model training method, definition detection method and device | |
| CN108960269B (en) | Feature acquisition method and device for data set and computing equipment | |
| CN111126028B (en) | Data processing method, device and equipment | |
| CN110956615A (en) | Image quality evaluation model training method and device, electronic equipment and storage medium | |
| CN115730605B (en) | Data analysis method based on multidimensional information | |
| CN115100739B (en) | Man-machine behavior detection method, system, terminal device and storage medium | |
| CN111444807A (en) | Target detection method, device, electronic equipment and computer readable medium | |
| CN109376736A (en) | A kind of small video target detection method based on depth convolutional neural networks | |
| CN116362782A (en) | User interest point identification method and system based on big data analysis | |
| KR20210152269A (en) | Method for learning of deep learning model and computing device for executing the method | |
| CN116109630A (en) | Image analysis method and system based on sensor acquisition and artificial intelligence | |
| CN116451081A (en) | Data drift detection method, device, terminal and storage medium | |
| CN116543538B (en) | Internet of things fire-fighting electrical early warning method and early warning system | |
| CN117591860A (en) | Data anomaly detection method and device | |
| CN116150186B (en) | Communication Internet of things management method and system based on big data | |
| CN115292571A (en) | App data acquisition method and system | |
| CN114638304A (en) | Training method of image recognition model, image recognition method and device | |
| CN111353349B (en) | Human body key point detection method and device, electronic equipment and storage medium | |
| CN115802355B (en) | Mobile Internet of things card management method and device and cloud platform | |
| CN116148806A (en) | Method and device for determining depth of target object and electronic equipment | |
| CN118312657B (en) | Knowledge base-based intelligent large model analysis recommendation system and method | |
| CN115984948B (en) | Face recognition method applied to temperature sensing and electronic equipment | |
| CN110837805B (en) | Method, device and equipment for measuring confidence of video tag and storage medium | |
| CN117292304B (en) | Multimedia data transmission control method and system |
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 |