CN113536139B - Content recommendation method and device based on interests, computer equipment and storage medium - Google Patents
Content recommendation method and device based on interests, computer equipment and storage medium Download PDFInfo
- Publication number
- CN113536139B CN113536139B CN202111077750.5A CN202111077750A CN113536139B CN 113536139 B CN113536139 B CN 113536139B CN 202111077750 A CN202111077750 A CN 202111077750A CN 113536139 B CN113536139 B CN 113536139B
- Authority
- CN
- China
- Prior art keywords
- sequence
- term
- short
- interest
- long
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a content recommendation method based on interests, which is applied to the technical field of informatization recommendation and is used for improving the accuracy of content recommendation of users. The method provided by the invention comprises the following steps: acquiring a historical browsing sequence of a user from a log file of a server, and classifying the historical browsing sequence into a long-term interest sequence and a short-term interest sequence according to a preset classification rule, wherein the historical browsing sequence comprises click behavior information of the user; inputting the long-term interest sequence into a preset graph neural network for feature extraction, and generating long-term prediction information; inputting the short-term interest sequence into a preset recurrent neural network for feature extraction, and generating short-term prediction information; and according to a preset verification method, performing result verification based on the long-term prediction information and the short-term prediction information to obtain a content recommendation result.
Description
Technical Field
The invention relates to the technical field of informatization recommendation, in particular to a content recommendation method and device based on interests, computer equipment and a storage medium.
Background
With the increasing development of computer technology and the internet, various information is increased explosively, users of the internet are surrounded by massive information, and a recommendation system is produced in order to more accurately mine the interests of the users. In the prior art, a recommendation system predicts the interest of a user according to the browsing history of the user, and recommends content meeting the interest of the user to the user so as to improve the browsing experience of the user.
The existing prediction method generally adopts a Recurrent Neural Network (RNN) to analyze the historical browsing behavior of a user, and in the browsing process, the interest points of the user are changed along with the analysis.
Disclosure of Invention
The invention provides a content recommendation method and device based on interests, computer equipment and a storage medium, which are used for improving the accuracy of content recommendation of a user based on the user interests.
An interest-based content recommendation method comprising:
acquiring a historical browsing sequence of a user from a log file of a server, and classifying the historical browsing sequence into a long-term interest sequence and a short-term interest sequence according to a preset classification rule, wherein the historical browsing sequence comprises click behavior information of the user;
inputting the long-term interest sequence into a preset graph neural network for feature extraction, and generating long-term prediction information;
inputting the short-term interest sequence into a preset recurrent neural network for feature extraction, and generating short-term prediction information;
and according to a preset verification method, performing result verification based on the long-term prediction information and the short-term prediction information to obtain a content recommendation result.
An interest-based content recommendation apparatus comprising:
the interest sequence classification module is used for acquiring a historical browsing sequence of a user from a log file of a server and classifying the historical browsing sequence into a long-term interest sequence and a short-term interest sequence according to a preset classification rule, wherein the historical browsing sequence comprises click behavior information of the user;
the long-term prediction information module is used for inputting the long-term interest sequence into a preset graph neural network for feature extraction to generate long-term prediction information;
the short-term prediction information module is used for inputting the short-term interest sequence into a preset recurrent neural network for feature extraction to generate short-term prediction information;
and the click prediction result module is used for carrying out result verification based on the long-term prediction information and the short-term prediction information according to a preset verification method to obtain a content recommendation result.
A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, the processor implementing the steps of the above-mentioned interest-based content recommendation method when executing the computer program.
A computer-readable storage medium, storing a computer program which, when executed by a processor, performs the steps of the above-described interest-based content recommendation method.
The invention provides an interest-based content recommendation method, a device, a computer device and a storage medium, which are characterized in that a historical browsing sequence of a user in a server is obtained, the historical browsing sequence is divided into a short-term interest sequence and a long-term interest sequence based on a change cycle of the user interest, interest characteristics of the long-term interest sequence are extracted through a graph neural network to generate long-term prediction information, interest characteristics of the short-term interest sequence are extracted through a cyclic neural network to generate short-term prediction information, result verification is carried out based on the long-term interest prediction and the short-term prediction information to obtain click prediction results, different feature extractors are respectively adopted to carry out feature extraction on the long-term interest sequence and the short-term interest sequence, the characteristics of the long-term interest and the short-term interest of the user are reserved, a content recommendation result is jointly generated by utilizing the long-term prediction information and the short-term prediction information, and the generated content recommendation result is fused with the change characteristics of the long-term and short-term user interest, therefore, the generated content recommendation result is closer to the interest of the user, and the accuracy of content recommendation is improved.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the description of the embodiments of the present invention will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art that other drawings can be obtained according to these drawings without inventive labor.
FIG. 1 is a diagram illustrating an application environment of a method for interest-based content recommendation according to an embodiment of the present invention;
FIG. 2 is a flow chart of a method for interest-based content recommendation in an embodiment of the invention;
FIG. 3 is a schematic diagram of an apparatus for recommending content based on interest according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of a computer device according to an embodiment of the invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The interest-based content recommendation method provided by the embodiment of the invention can be applied to an application environment as shown in fig. 1, wherein the terminal device communicates with the server through a network. The terminal device may be, but is not limited to, various personal computers, notebook computers, smart phones, tablet computers, and portable wearable devices. The server may be implemented as a stand-alone server or as a server cluster consisting of a plurality of servers.
The system framework 100 may include terminal devices, networks, and servers. The network serves as a medium for providing a communication link between the terminal device and the server. The network may include various connection types, such as wired, wireless communication links, or fiber optic cables, to name a few.
A user may use a terminal device to interact with a server over a network to receive or send messages or the like.
The terminal devices 101, 102, 103 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, E-book readers, MP3 players (Moving Picture E interface displays properties Group Audio Layer III, motion Picture experts compress standard Audio Layer 3), MP4 (Moving Picture experts Group Audio Layer IV, motion Picture experts compress standard Audio Layer 4) players, laptop portable computers, desktop computers, and the like.
The server 105 may be a server providing various services, such as a background server providing support for pages displayed on the terminal devices 101, 102, 103.
It should be noted that the interest-based content recommendation method provided by the embodiment of the present invention is executed by a server, and accordingly, an interest-based content recommendation apparatus is disposed in the server.
It should be understood that the number of the terminal devices, the networks, and the servers in fig. 1 is only illustrative, and any number of the terminal devices, the networks, and the servers may be provided according to implementation requirements, and the terminal devices in the embodiment of the present invention may specifically correspond to an application system in actual production.
In an embodiment, as shown in fig. 2, a method for recommending content based on interest is provided, which is described by taking the method as an example applied to the server in fig. 1, and includes the following steps S201 to S204.
S201, obtaining a historical browsing sequence of a user from a log file of a server, and classifying the historical browsing sequence into a long-term interest sequence and a short-term interest sequence according to a preset classification rule, wherein the historical browsing sequence comprises click behavior information of the user.
The server refers to a computer device for storing browsing information of a user, and the user stores browsing behavior in a log file when browsing a page. And acquiring a historical browsing sequence of the user from a log file of the server, wherein the clicking behavior information of the user in the historical browsing sequence comprises a clicking object and clicking time of the user in the process of browsing the internet content in a historical period. The click behavior information refers to the object of the click behavior and the time when the click behavior occurs. For example, when the user browses a web page of a shopping website, the generated historical browsing sequence is { product a-23: 00PM, product B-23: 15PM }.
And classifying the historical browsing sequences into long-term interest sequences and short-message interest sequences according to a preset time period dividing method and time distribution of the historical browsing sequences.
And taking all historical browsing sequences of a certain time period as long-term interest sequences, further dividing the time period, and taking the last divided time period as short-term interest sequences. For example, if the time period of browsing the web page of the shopping website by the user a on a certain day is morning, noon and evening, and the current time period is the evening time period of the day, the historical browsing sequence generated on the day is used as a long-term interest sequence, and the historical browsing sequence of the current time period, that is, the evening time period of the day is used as a short-term interest sequence.
S202, inputting the long-term interest sequence into a preset graph neural network for feature extraction, and generating long-term prediction information.
The graph neural network expresses the long-term interest sequence of the user in a graph form, and the graph stores the overall relevance of each click object in the long-term interest sequence. And each click object is represented by a node in the graph, and the relation between each click object is represented by an edge, wherein the edge between the click objects represents the interest conversion probability of each click object and the attention weight of the click object.
The interest conversion probability and the attention weight between the click objects are generated through a graph attention module in a graph neural network, the graph attention module calculates the similarity between the click objects, and the similarity of each click object is obtained according to the calculation of the feature vector of each click object.
And generating a similarity matrix of the long-term interest sequence according to the similarity between each click object, wherein the lower the similarity between the two click objects is, the greater the corresponding attention weight is. And (4) always processing the special benefits in the map through a map gating module to obtain long-term prediction information.
The long-term prediction information includes interest conversion probability of each click object in the long-term interest sequence and attention weight of each click object.
And S203, inputting the short-term interest sequence into a preset recurrent neural network for feature extraction, and generating short-term prediction information.
The short-term interest sequence of the user is input into the recurrent neural network to carry out feature extraction, wherein the recurrent neural network comprises an input gate, a forgetting gate and an output gate, and the short-term interest sequence of the user is input into the recurrent neural network to carry out feature extraction so as to obtain short-term prediction information.
And S204, according to a preset verification method, performing result verification based on the long-term prediction information and the short-term prediction information to obtain a content recommendation result.
Generating a real result by generating an antagonistic neural network, comparing the long-term prediction information with the real result, and adjusting the parameters of the neural network when the results of the two are different; and comparing the short-term prediction result with the real result, if the short-term prediction result is different from the real result, adjusting parameters of the graph neural network, and generating a content recommendation result when the long-term prediction result is the same as the short-term prediction result, wherein the content recommendation result comprises a prediction result of the user clicking behavior, and each prediction result is used as a content recommendation result.
Furthermore, the model parameters of the graph neural network and the cyclic neural network are confronted to continuously generate a real prediction sequence, wherein the length of the prediction sequence is N, the first N-1 click sequence of the prediction sequence is a fixed real value, the output of the Nth value is simulated through the real value, a loss function is obtained according to the generated prediction sequence, and the loss function can be expressed according to the following formula:
the losses are calculated for the graph neural network using the above formula. Wherein
The representation is a neural network function of the generator graph,
expressed as a reward function, the reward function is expressed according to the following formula:
wherein the content of the first and second substances,
which is indicative of the current sequence,
representing the next click object of the current sequence prediction,
a set of next click objects representing the current sequence,
the corresponding bias of the function is represented. By gated recurrent neural networks
And performing iterative training to obtain a reward function.
The method, the device, the computer equipment and the storage medium for recommending the content based on the interest, which are provided by the embodiment of the invention, have the advantages that the historical browsing sequence of the user in the server is obtained, the historical browsing sequence is divided into a short-term interest sequence and a long-term interest sequence based on the change cycle of the user interest, the interest characteristics of the long-term interest sequence are extracted through a graph neural network to generate long-term prediction information, the interest characteristics of the short-term interest sequence are extracted through a circulating neural network to generate short-term prediction information, the result verification is carried out based on the long-term interest prediction information and the short-term prediction information at the same time to obtain the click prediction result, the long-term interest sequence and the short-term interest sequence are respectively extracted through different feature extractors, the long-term interest characteristics and the short-term interest characteristics of the user are reserved, the content recommendation result is jointly generated by utilizing the long-term prediction information and the short-term prediction information, and the generated content recommendation result is fused with the change characteristics of the long-term interest and the short-term interest of the user, therefore, the generated content recommendation result is closer to the interest of the user, and the accuracy of content recommendation is improved.
In this embodiment, as an optional implementation manner, in step S201, a historical browsing sequence of the user is obtained from a log file of the server, and the historical browsing sequence is classified into a long-term interest sequence and a short-term interest sequence according to a preset classification rule, as shown below.
And S2011, acquiring click behavior information in the historical browsing sequence as a click sequence, wherein the click behavior information comprises an object of click behavior and time of click behavior.
And S2012, taking the click sequence as a long-term interest sequence.
S2013, segmenting the click sequences according to the preset time periods to obtain a plurality of short-term click sequences, and taking the short-term sequence of the last time period as a short-term interest sequence.
Further, in this embodiment, the history browsing sequence needs to be processed.
The method comprises the steps that click sequence information of all users of the shopping software is contained in an online shopping scene, a research area is set to be a product with a high part of click frequency, products with the click frequency lower than the set click frequency are filtered, then click sequences of the users are divided according to time, and click sequence information of all the users is obtained, for example: { apple, orange, pear } a click sequence of length 3.
And then, preprocessing the click sequence information to strengthen the click sequence information to obtain data, wherein in the process, the sequence S = { S1, S2,. sn } is changed into ([ S1], S2), ([ S1, S2], S3). ([ S1, S2,. sn-1 ], sn), so that the data with the original length n can be expanded to n times of the original length to better learn the time sequence characteristics of the user, and a final historical browsing sequence is obtained.
In the embodiment, the click sequence information of the user is processed to obtain the historical browsing sequence, and the useless information is screened to ensure the data availability of the historical browsing sequence. The long-term interest sequence and the short-term interest sequence are obtained by classifying the historical browsing sequences, and the features of the long-term interest sequence and the short-term interest sequence are extracted in different feature extraction modes, so that the long-term interest and the short-term interest of the user are reserved, and the accuracy of content recommendation is improved.
In this embodiment, as an optional implementation manner, in step S202, the long-term interest sequence is input to a preset neural network of a graph for feature extraction, and the step of generating long-term prediction information is as follows.
S2021, extracting the click behavior characteristics of the long-term interest sequence through a preset graph neural network, and taking the click behavior sequence of the long-term interest sequence as a long-term characteristic sequence.
The click behavior features are extracted through a graph neural network to generate a graph, each node in the graph is expressed as a click object, and edges among all the click objects comprise the probability of interest conversion among entities, entity attention weights and gating weights. And simultaneously, the final click of the user trains the capability of the neural network for extracting the user characteristics, and the process is to calculate by using the current sequence relation parameter, the attention parameter and the gating parameter through each neuron of the neural network to obtain a long-term characteristic sequence.
S2022, generating an attention weight and a probability value of each sequence value of the long-term feature sequence through a preset attention module, and generating long-term prediction information of the long-term feature sequence based on each sequence value and the corresponding probability value.
Wherein, the similarity between each click object is calculated through a graph attention module in the graph neural network. The functions of the graph attention module are: and (3) feature amplification is carried out on the main features in the historical click messages of the users by calculating the similarity among the products, wherein the main features are expressed as the change of the main interest in the current entity set. Constructing a similarity matrix between each clicked object, and if the similarity between the clicked objects is lower and the features of the clicked objects are more prominent, giving higher initial weight and optimized weight to the clicked objects for feature amplification, so as to avoid rapid smoothing caused by continuous iteration; and weakening the weight of the click objects appearing in the click object set for many times, giving lower initial weight and optimized weight, avoiding repeatedly calculating the characteristics of the same entity for many times, and preventing the final long-term prediction information from excessively deviating to the secondary interest.
And then screening the click objects with low similarity in the map by using the map gating module, thereby avoiding the influence on the long-term prediction information.
In this embodiment, the long-term prediction information is obtained by extracting the interest features in the long-term interest sequence through the graph neural network, wherein the features of each click object in the long-term prediction information are processed, so that the interest of each click object in the long-term interest sequence is based on a steady state, and the long-term interest of the user is more accurately reflected.
In this embodiment, as an optional implementation manner, in step S203, the short-term interest sequence is input to a preset recurrent neural network for feature extraction, and the step of generating short-term prediction information is as follows.
S2031, extracting click behavior characteristics of the short-term interest sequence through a preset recurrent neural network, and taking the click behavior characteristics of the short-term interest sequence as a short-term characteristic sequence.
The preset cyclic neural network consists of an input gate, a forgetting gate and an output gate.
The input gate is used for fusing input short-term interest sequences and is expressed according to the following formula:
wherein the content of the first and second substances,
a current click object representing the input sequence,
representing the first t-1 set of click objects.
And
the corresponding weight is represented by a weight that is,
representing the corresponding bias coefficient.
The forgetting gate controls how much past memory features are forgotten. The forgetting gate can be expressed according to the following formula:
wherein the content of the first and second substances,
a current click object representing the input sequence,
representing the first t-1 set of click objects.
And
the corresponding weight is represented by a weight that is,
representing the corresponding bias coefficient.
The memory characteristic of the output gate for controlling the output can be expressed according to the following formula:
wherein the content of the first and second substances,
a current click object representing the input sequence,
representing the first t-1 set of click objects.
And
the corresponding weight is represented by a weight that is,
representing the corresponding bias coefficient.
S2032, calculating the attention weight and the probability value of each sequence value in the short-term characteristic sequence, and generating short-term prediction information of the short-term characteristic sequence based on each sequence value and the corresponding probability value.
And generating corresponding attention weight and probability value for the sequence value in the short-term characteristic sequence through the neural network to obtain short-term prediction information.
In the embodiment, the short-term prediction information reflecting the short-term interest of the user is obtained by extracting the features in the short-term interest sequence through the recurrent neural network, so that the short-term interest is reflected in the finally generated content recommendation result, the accuracy of content recommendation is improved, the browsing experience of the user is improved,
in this embodiment, as an optional implementation manner, in step S204, according to a preset verification method, a result prediction is performed based on the long-term prediction information and the short-term prediction information, and steps of obtaining a content recommendation result are as follows.
And S2041, inputting the long-term prediction information and the short-term prediction information into a preset generation countermeasure network for result prediction to obtain a content recommendation result.
The generated countermeasure network generates countermeasures through model parameters of the graph neural network and the circulation neural network, click prediction results are generated through the generated countermeasure network, and the optimal click prediction results are obtained through continuous back propagation and continuous optimization and serve as content prediction results.
The training process for generating the countermeasure network is a cooperative training process of the graph neural network and the recurrent neural network, and the parameters are alternately trained and optimized, wherein in order to make the loss function and the reward function of the generator and the arbiter finally converge at one point, the prediction results generated by the two are the same.
In this embodiment, the graph neural network and the recurrent neural network perform iterative training on the click prediction result, an optimized click recommendation result is generated and is used as a content recommendation result, the content recommendation result includes the short-term interest and the long-term interest of the user, and the content recommendation result generated on the basis is closer to the click behavior of the user, so that the accuracy of generating the content recommendation information can be improved.
In this embodiment, in step S2041, the generation countermeasure network is composed of a generation countermeasure module and a comparison module, and further includes the following steps, as shown below.
And S1, inputting the long-term prediction information and the short-term prediction information into the generation countermeasure module to obtain an initial prediction result.
Wherein, by generating the countermeasure network, an initial prediction sequence is generated based on the long-term prediction information and the short-term prediction information, the initial prediction sequence including a prediction sequence in the long-term prediction information and the short-term prediction information.
And S2, matching and judging the initial prediction result and the real click sequence through a comparison module to obtain a judgment result, and taking the judgment result matched with the real click sequence as a content recommendation result.
The method comprises the steps of comparing a prediction sequence in an initial prediction result with a real click sequence, setting a discriminator, carrying out parameter adjustment on a generated countermeasure network by the discriminator according to the comparison result, enabling the generated countermeasure network to generate a more accurate prediction sequence, and finally generating an optimal prediction sequence as a content recommendation result through iterative adjustment.
In this embodiment, by comparing the initial prediction result with the real click sequence and adjusting the parameters for generating the countermeasure network according to the comparison result, the generated content recommendation result is more accurate, and the accuracy of the recommendation information generated when the user is recommended is improved.
It should be understood that, the sequence numbers of the steps in the foregoing embodiments do not imply an execution sequence, and the execution sequence of each process should be determined by its function and inherent logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.
In one embodiment, an interest-based content recommendation apparatus is provided, and the interest-based content recommendation apparatus corresponds to the interest-based content recommendation method in the above embodiments one to one. As shown in fig. 3, the interest-based content recommendation apparatus includes an interest sequence classification module 31, a long-term prediction information module 32, a short-term prediction information module 33, and a click prediction result module 34.
The interest sequence classification module 31 is configured to acquire a historical browsing sequence of the user from a log file of the server, and classify the historical browsing sequence into a long-term interest sequence and a short-term interest sequence according to a preset classification rule, where the historical browsing sequence includes click behavior information of the user.
And the long-term prediction information module 32 is used for inputting the long-term interest sequence into a preset graph neural network for feature extraction, and generating long-term prediction information.
And a short-term prediction information module 33, configured to input the short-term interest sequence into a preset recurrent neural network to perform feature extraction, so as to generate short-term prediction information.
And the click prediction result module 34 is configured to perform result verification based on the long-term prediction information and the short-term prediction information according to a preset verification method to obtain a content recommendation result.
In the present embodiment, the interest sequence classification model 31 includes the following units.
And the click sequence unit is used for acquiring click behavior information in the historical browsing sequence as a click sequence, wherein the click behavior information comprises an object of click behavior and time of click behavior.
And the long-term interest sequence unit is used for taking the click sequence as the long-term interest sequence.
And the short-term interest sequence unit is used for segmenting the click sequences according to a preset time period to obtain a plurality of short-term click sequences, and taking the short-term sequence of the last time period as the short-term interest sequence.
In the present embodiment, the long-term prediction information module 32 includes the following units.
And the long-term feature extraction unit is used for extracting the click behavior features of the long-term interest sequence through a preset graph neural network, and taking the click behavior sequence of the long-term interest sequence as a long-term feature sequence.
And the long-term prediction information unit is used for generating an attention weight and a probability value of each sequence value of the long-term characteristic sequence through a preset attention module, and generating long-term prediction information of the long-term characteristic sequence based on each sequence value and the corresponding probability value.
In the present embodiment, the short-term prediction information module 33 includes the following units.
And the short-term feature extraction unit is used for extracting the click behavior features of the short-term interest sequence through a preset recurrent neural network and taking the click behavior features of the short-term interest sequence as a short-term feature sequence.
And the short-term prediction information unit is used for calculating the attention weight and the probability value of each sequence value in the short-term characteristic sequence and generating the short-term prediction information of the short-term characteristic sequence based on each sequence value and the corresponding probability value.
In this embodiment, the click predictor module 34 includes the following elements.
And the content recommendation result is used for inputting the long-term prediction information and the short-term prediction information into a preset generation countermeasure network to perform result prediction so as to obtain a content recommendation result.
Further, the following units are also included.
And the initial prediction result unit is used for inputting the long-term prediction information and the short-term prediction information into the generation countermeasure module to obtain an initial prediction result.
And the content recommendation result unit is used for matching and judging the initial prediction result and the real click sequence through the comparison module to obtain a judgment result, and taking the judgment result matched with the real click sequence as a content recommendation result.
Wherein the meaning of "first" and "second" in the above modules/units is only to distinguish different modules/units, and is not used to define which module/unit has higher priority or other defining meaning. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or modules is not necessarily limited to those steps or modules explicitly listed, but may include other steps or modules not explicitly listed or inherent to such process, method, article, or apparatus, and such that a division of modules presented in this application is merely a logical division and may be implemented in a practical application in a further manner.
For specific limitations of the interest-based content recommendation device, reference may be made to the above limitations of the interest-based content recommendation method, which are not described herein again. The respective modules in the interest-based content recommendation apparatus may be wholly or partially implemented by software, hardware, and a combination thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.
In one embodiment, a computer device is provided, which may be a server, the internal structure of which may be as shown in fig. 4. The computer device includes a processor, a memory, a network interface, and a database connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The database of the computer device is used for storing data involved in the interest-based content recommendation method. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a method of interest-based content recommendation.
In one embodiment, a computer device is provided, which includes a memory, a processor, and a computer program stored on the memory and executable on the processor, and the processor executes the computer program to implement the steps of the interest-based content recommendation method in the above embodiments, such as the steps S201 to S204 shown in fig. 2 and other extensions of the method and related steps. Alternatively, the processor, when executing the computer program, implements the functions of the modules/units of the interest-based content recommendation apparatus in the above-described embodiments, such as the functions of the interest sequence classification module 31, the long-term prediction information module 32, the short-term prediction information module 33, and the click prediction result module 34 shown in fig. 3. To avoid repetition, further description is omitted here.
The Processor may be a Central Processing Unit (CPU), other general purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), an off-the-shelf Programmable Gate Array (FPGA) or other Programmable logic device, discrete Gate or transistor logic device, discrete hardware component, etc. The general purpose processor may be a microprocessor or the processor may be any conventional processor or the like which is the control center for the computer device and which connects the various parts of the overall computer device using various interfaces and lines.
The memory may be used to store the computer programs and/or modules, and the processor may implement various functions of the computer device by running or executing the computer programs and/or modules stored in the memory and invoking data stored in the memory. The memory may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required by at least one function (such as a sound playing function, an image playing function, etc.), and the like; the storage data area may store data (such as audio data, video data, etc.) created according to the use of the cellular phone, etc.
The memory may be integrated in the processor or may be provided separately from the processor.
In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored, which, when being executed by a processor, implements the steps of the interest-based content recommendation method in the above-described embodiments, such as the steps S201 to S204 shown in fig. 2 and extensions of other extensions and related steps of the method. Alternatively, the computer program when executed by the processor implements the functions of the modules/units of the interest-based content recommendation apparatus in the above-described embodiments, such as the functions of the interest sequence classification module 31, the long-term prediction information module 32, the short-term prediction information module 33, and the click prediction result module 34 shown in fig. 3. To avoid repetition, further description is omitted here.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in the embodiments provided herein may include non-volatile and/or volatile memory, among others. Non-volatile memory can include read-only memory (ROM), Programmable ROM (PROM), Electrically Programmable ROM (EPROM), Electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDRSDRAM), Enhanced SDRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), Rambus Direct RAM (RDRAM), direct bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
It will be apparent to those skilled in the art that, for convenience and brevity of description, only the above-mentioned division of the functional units and modules is illustrated, and in practical applications, the above-mentioned function distribution may be performed by different functional units and modules according to needs, that is, the internal structure of the apparatus is divided into different functional units or modules to perform all or part of the above-mentioned functions.
The above-mentioned embodiments are only used for illustrating the technical solutions of the present invention, and not for limiting the same; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; such modifications and substitutions do not substantially depart from the spirit and scope of the embodiments of the present invention, and are intended to be included within the scope of the present invention.
Claims (7)
1. An interest-based content recommendation method, comprising:
acquiring a historical browsing sequence of a user from a log file of a server, and classifying the historical browsing sequence into a long-term interest sequence and a short-term interest sequence according to a preset classification rule, wherein the historical browsing sequence comprises click behavior information of the user;
inputting the long-term interest sequence into a preset graph neural network for feature extraction, and generating long-term prediction information;
inputting the short-term interest sequence into a preset recurrent neural network for feature extraction, and generating short-term prediction information;
inputting the long-term prediction information and the short-term prediction information into a preset generation countermeasure network for training to obtain a reward function, wherein the generation countermeasure network comprises a generator and a discriminator, the generator is constructed based on a graph neural network, and the discriminator is constructed based on a cyclic neural network;
and carrying out weight adjustment on the long-term prediction information and the short-term prediction information through the reward function, searching for an interest sequence to obtain an initial recommendation result, and carrying out matching judgment on the initial recommendation result through a comparison module to obtain a content recommendation result.
2. The interest-based content recommendation method according to claim 1, wherein the step of obtaining the historical browsing sequence of the user from the log file of the server and classifying the historical browsing sequence into a long-term interest sequence and a short-term interest sequence according to a preset classification rule comprises:
acquiring click behavior information in the historical browsing sequence as a click sequence, wherein the click behavior information comprises an object of a click behavior and time of the click behavior;
taking the click sequence as the long-term interest sequence;
and segmenting the click sequence according to a preset time period to obtain a plurality of short-term click sequences, and taking the short-term sequence of the last time period as a short-term interest sequence.
3. The interest-based content recommendation method according to claim 1, wherein the step of inputting the long-term interest sequence into a preset neural network for feature extraction and generating long-term prediction information comprises:
extracting the click behavior characteristics of the long-term interest sequence through the preset graph neural network, and taking the click behavior sequence of the long-term interest sequence as a long-term characteristic sequence;
and generating attention weight and probability value of each sequence value of the long-term characteristic sequence through a preset attention module, and generating long-term prediction information of the long-term characteristic sequence based on each sequence value and the corresponding probability value.
4. The interest-based content recommendation method according to claim 1, wherein the step of inputting the short-term interest sequence into a preset recurrent neural network for feature extraction and generating short-term prediction information comprises:
extracting the click behavior characteristics of the short-term interest sequence through the preset recurrent neural network, and taking the click behavior characteristics of the short-term interest sequence as a short-term characteristic sequence;
and calculating the attention weight and the probability value of each sequence value in the short-term feature sequence, and generating short-term prediction information of the short-term feature sequence based on each sequence value and the corresponding probability value.
5. The interest-based content recommendation method according to claim 1, wherein the step of performing weight adjustment on the long-term prediction information and the short-term prediction information through the reward function to obtain an initial recommendation result, and performing matching judgment on the initial recommendation result through a comparison module to obtain a content recommendation result comprises:
adjusting the click sequence weight in the long-term prediction information through the reward function to generate an initial recommendation result;
and matching and judging the initial prediction result and the real click sequence through the comparison module to obtain a judgment result, and taking the judgment result matched with the real click sequence as the content recommendation result.
6. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that the processor implements the steps of the interest-based content recommendation method according to any one of claims 1 to 5 when executing the computer program.
7. A computer-readable storage medium, in which a computer program is stored, which, when being executed by a processor, carries out the steps of the interest-based content recommendation method according to any one of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111077750.5A CN113536139B (en) | 2021-09-15 | 2021-09-15 | Content recommendation method and device based on interests, computer equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111077750.5A CN113536139B (en) | 2021-09-15 | 2021-09-15 | Content recommendation method and device based on interests, computer equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113536139A CN113536139A (en) | 2021-10-22 |
| CN113536139B true CN113536139B (en) | 2022-02-11 |
Family
ID=78093227
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111077750.5A Active CN113536139B (en) | 2021-09-15 | 2021-09-15 | Content recommendation method and device based on interests, computer equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113536139B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113886721B (en) * | 2021-12-06 | 2022-02-18 | 湖南工商大学 | Personalized interest point recommendation method and device, computer equipment and storage medium |
| CN115062230B (en) * | 2022-08-16 | 2023-01-06 | 深圳市海清视讯科技有限公司 | Data processing method, apparatus, device, medium, and program product |
| CN117082020B (en) * | 2023-10-13 | 2023-12-29 | 湖南双鸿科技有限公司 | Automatic forwarding method for friend circle based on instant messaging software |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101339563B (en) * | 2008-08-15 | 2010-06-02 | 北京航空航天大学 | Interest model update method facing to odd discovery recommendation |
| US10458990B1 (en) * | 2015-03-06 | 2019-10-29 | Scanit Technologies, Inc. | Spore state discrimination |
| US20190258984A1 (en) * | 2018-02-19 | 2019-08-22 | Microsoft Technology Licensing, Llc | Generative adversarial networks in predicting sequential data |
| CN110457589B (en) * | 2019-08-19 | 2020-05-12 | 上海新共赢信息科技有限公司 | Vehicle recommendation method, device, equipment and storage medium |
| US11314529B2 (en) * | 2020-01-21 | 2022-04-26 | Walmart Apollo, Llc | Dynamical switching between long-term and short-term rewards |
| CN111915395A (en) * | 2020-07-07 | 2020-11-10 | 云境商务智能研究院南京有限公司 | Travel bag recommendation method based on multi-view attention mechanism |
-
2021
- 2021-09-15 CN CN202111077750.5A patent/CN113536139B/en active Active
Non-Patent Citations (1)
| Title |
|---|
| "MN··HDRM: 长短兴趣多神经网络混合动态推荐模型";冯永等;《计算机学报》;20190131;第16-28页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113536139A (en) | 2021-10-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11893071B2 (en) | Content recommendation method and apparatus, electronic device, and storage medium | |
| CN113536139B (en) | Content recommendation method and device based on interests, computer equipment and storage medium | |
| CN109543109B (en) | Recommendation algorithm integrating time window technology and scoring prediction model | |
| CN111382361B (en) | Information pushing method, device, storage medium and computer equipment | |
| CN110598845A (en) | Data processing method, data processing device, computer equipment and storage medium | |
| CN110941764A (en) | Object recommendation method and device, computer equipment and storage medium | |
| CN112765480B (en) | Information pushing method and device and computer readable storage medium | |
| CN109376237A (en) | Prediction technique, device, computer equipment and the storage medium of client's stability | |
| CN114528844A (en) | Intention recognition method and device, computer equipment and storage medium | |
| JP2024503774A (en) | Fusion parameter identification method and device, information recommendation method and device, parameter measurement model training method and device, electronic device, storage medium, and computer program | |
| CN112085565A (en) | Deep learning-based information recommendation method, device, equipment and storage medium | |
| CN114036398B (en) | Content recommendation and ranking model training method, device, equipment and storage medium | |
| Wang et al. | A personalized electronic movie recommendation system based on support vector machine and improved particle swarm optimization | |
| CN113343091A (en) | Industrial and enterprise oriented science and technology service recommendation calculation method, medium and program | |
| CN114399396A (en) | Insurance product recommendation method and device, computer equipment and storage medium | |
| Chen et al. | A new approach for mobile advertising click-through rate estimation based on deep belief nets | |
| Zhang et al. | Recurrent convolutional neural network for session-based recommendation | |
| CN112131261A (en) | Community query method and device based on community network and computer equipment | |
| CN116910357A (en) | Data processing method and related device | |
| CN114266324B (en) | Model visualization modeling method and device, computer equipment and storage medium | |
| CN114491093B (en) | Multimedia resource recommendation and object representation network generation method and device | |
| WO2022213662A1 (en) | Application recommendation method and system, terminal, and storage medium | |
| CN114428838A (en) | Content recall method and device, computer equipment and storage medium | |
| CN115618950A (en) | Data processing method and related device | |
| Shilin | User model-based personalized recommendation algorithm for news media education resources |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |