CN115730217A - Model training method, material recalling method and device - Google Patents
Model training method, material recalling method and device Download PDFInfo
- Publication number
- CN115730217A CN115730217A CN202211604889.5A CN202211604889A CN115730217A CN 115730217 A CN115730217 A CN 115730217A CN 202211604889 A CN202211604889 A CN 202211604889A CN 115730217 A CN115730217 A CN 115730217A
- Authority
- CN
- China
- Prior art keywords
- user
- target
- vector
- feature
- model
- 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
Images
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The application discloses a model training method, a material recalling method and a material recalling device, wherein the training method comprises the following steps: acquiring a plurality of user feature sets corresponding to a target user within first preset time and a material set consisting of materials clicked by the user within the first preset time, wherein the user feature sets are sets of user features corresponding to the target user each time the material is clicked; determining a feature vector corresponding to the user feature set based on the user feature set; performing iterative training on a target model to be trained through a plurality of characteristic vectors, inputting the characteristic vectors into the target model to be trained in each iterative training, and acquiring interest preference vectors which are output by the target model to be trained and correspond to target users; determining an estimated click probability of a target material clicked by a target user within a first preset time based on the interest preference vector and a material vector matrix constructed according to the material set; and under the condition that the estimated click probability reaches a target value, obtaining a trained target model.
Description
Technical Field
The application relates to the field of information recall, in particular to a model training method, a material recall method and a material recall device.
Background
In a social scene, the personalized recommendation system is responsible for pushing microblog messages (namely materials) of all bloggers to interested users in a message pushing mode, so that the users can acquire the interested information in time. The personalized recommendation system comprises a recall module, and the recall module screens out the possible interesting blog articles of thousands of users from the million-level candidate blog articles library by applying a plurality of recall algorithms to form a recall blog library. Because the user collaborative filtering algorithm can calculate the similarity of the users through the historical behaviors of the users and recommend materials liked by similar users for the users, the user collaborative filtering algorithm is widely applied to recall modules of various personalized recommendation systems.
In the related art, the calculation steps of the user collaborative filtering algorithm are as follows: the method comprises the steps of constructing a behavior matrix based on user Identification (ID) and material IDs, decomposing the behavior matrix into a user matrix representing interest preference of users and a material matrix representing material attributes, wherein in this case, the interest preference of each user can be represented by a vector corresponding to the user in the user matrix, grouping the users based on the vectors corresponding to the users, counting favorite materials of the same group of users, and recalling the materials for other users of the same group who do not distribute the materials. However, the above scheme determines the interest preference of the user based on the user ID and the material ID, and has a problem of low accuracy in representing the interest preference of the user.
Disclosure of Invention
The application discloses a model training method, a material recalling method and a material recalling device, which are used for solving the problem of low accuracy in expressing interest and preference of a user in the related art.
In order to solve the above problems, the following technical solutions are adopted in the present application:
in a first aspect, an embodiment of the present application provides a method for training a model, including: the method comprises the steps of obtaining a plurality of user feature sets corresponding to a target user within first preset time and a material set consisting of materials clicked by the user within the first preset time, wherein the user feature sets are sets of user features corresponding to the target user each time the target user clicks the materials; determining a feature vector corresponding to the user feature set based on the user feature set; performing iterative training on a target model to be trained through a plurality of feature vectors, wherein the feature vectors are input into the target model to be trained in each iterative training, and an interest preference vector corresponding to the target user and output by the target model to be trained is obtained; determining the estimated click probability of the target material clicked by the target user within the first preset time based on the interest preference vector and a material vector matrix constructed according to the material set; and obtaining a trained target model under the condition that the estimated click probability reaches a target value.
In a second aspect, an embodiment of the present application provides a method for recalling a material, including: respectively acquiring user feature sets corresponding to a plurality of users; determining a feature vector corresponding to the user feature set based on the user feature set; respectively inputting each feature vector in the feature vectors into a target model, and acquiring a plurality of interest preference vectors corresponding to the user and output by the target model; determining at least one user group by clustering a plurality of the interest preference vectors; respectively acquiring the times of clicks of users in a target user group on a plurality of materials within second preset time, wherein the target user group is any user group in the at least one user group; determining a hot spot material list corresponding to the target user group based on the click times; recalling materials for users in the target user group based on the hot spot material list; wherein the target model is obtained by training according to the training method of the model of the first aspect.
In a third aspect, an embodiment of the present application provides a training apparatus for a model, including: the system comprises an acquisition module, a storage module and a display module, wherein the acquisition module is used for acquiring a plurality of user feature sets corresponding to a target user within first preset time and a material set consisting of materials clicked by the user within the first preset time, and the user feature sets are sets of user features corresponding to the target user each time the material is clicked; the first determination module is used for determining a feature vector corresponding to the user feature set based on the user feature set; the training module is used for performing iterative training on a target model to be trained through a plurality of feature vectors, wherein the feature vectors are input into the target model to be trained in each iterative training, and an interest preference vector which is output by the target model to be trained and corresponds to the target user is obtained; the second determination module is used for determining the estimated click probability of the target material clicked by the target user within the first preset time based on the interest preference vector and a material vector matrix constructed according to the material set; and the obtaining module is used for obtaining the trained target model under the condition that the estimated click probability reaches the target value.
In a fourth aspect, an embodiment of the present application provides a material recalling apparatus, including: the first acquisition module is used for respectively acquiring user feature sets corresponding to a plurality of users; the first determination module is used for determining a feature vector corresponding to the user feature set based on the user feature set; a second obtaining module, configured to input each of the feature vectors into a target model, and obtain a plurality of interest preference vectors corresponding to the user and output by the target model; a second determining module for determining at least one user group by clustering a plurality of the interest preference vectors; a third obtaining module, configured to respectively obtain click times of multiple materials by users in a target user group within a second preset time, where the target user group is any user group in the at least one user group; a third determining module, configured to determine, based on the number of clicks, a hot spot material list corresponding to the target user group; a recall module, configured to recall the materials for the users in the target user group based on the hot spot material list; wherein the target model is obtained by training according to the training method of the model of the first aspect.
In a fifth aspect, embodiments of the present application provide an electronic device, which includes a processor and a memory, where the memory stores a program or instructions executable on the processor, and the program or instructions, when executed by the processor, implement the steps of the method according to the first aspect or the second aspect.
In a sixth aspect, embodiments of the present application provide a readable storage medium, on which a program or instructions are stored, which when executed by a processor implement the steps of the method according to the first or second aspect.
In a seventh aspect, an embodiment of the present application provides a chip, where the chip includes a processor and a communication interface, where the communication interface is coupled to the processor, and the processor is configured to execute a program or instructions to implement the method according to the first aspect or the second aspect.
In an eighth aspect, embodiments of the present application provide a computer program product, stored on a storage medium, for execution by at least one processor to implement a method according to the first or second aspect.
The embodiment of the application provides a model training method, a plurality of user feature sets corresponding to a target user in first preset time and a material set consisting of materials clicked by the user in the first preset time are obtained, feature vectors corresponding to the user feature sets are determined based on the user feature sets, the target model to be trained is subjected to iterative training through the feature vectors, then, the click probability of the target materials clicked by the target user in the first preset time is determined based on interest preference vectors output by the target model to be trained and corresponding to the target user and a material vector matrix constructed according to the material set, the trained target model is obtained under the condition that the estimated click probability reaches a target value, the trained target model can output the interest preference vectors corresponding to the user based on the user feature sets, and the user interest preference representation accuracy is high.
The embodiment of the application provides a material recalling method, which includes the steps of respectively obtaining user feature sets corresponding to users, determining feature vectors corresponding to the user feature sets based on the user feature sets, respectively inputting each feature vector in a plurality of feature vectors into a target model, obtaining a plurality of interest preference vectors corresponding to the users and output by the target model, then determining at least one user group by clustering the interest preference vectors, respectively obtaining the number of times of clicking on the plurality of materials by the users of the target user group within a second preset time, determining a hot point material list corresponding to the target user group based on the number of times of clicking on the plurality of materials by the users of the target user group, and recalling the materials for the users in the target user group based on the hot point material list. According to the material recalling method, the relevance between the materials recalled by the users and the interest preference of the users is high, the users are clustered into at least one user group, and then the materials are recalled for the users in the user group based on the interest preference of each user group, so that the calculated amount of online service can be reduced.
Drawings
FIG. 1 is a schematic flow chart illustrating a method for training a model according to an embodiment of the present disclosure;
FIG. 2 is a schematic flow chart illustrating a method for recalling materials according to an embodiment of the present disclosure;
FIG. 3 is a schematic structural diagram of a training apparatus for a model according to an embodiment of the present disclosure;
FIG. 4 is a schematic structural diagram of a material recall device according to an embodiment of the present disclosure;
fig. 5 is a schematic structural diagram of an electronic device disclosed in an embodiment of the present application.
Detailed Description
The technical solutions in the embodiments of the present application will be described clearly below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are some, but not all, embodiments of the present application. All other embodiments that can be derived by one of ordinary skill in the art from the embodiments given herein are intended to be within the scope of the present disclosure.
The terms first, second and the like in the description and in the claims of the present application are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It will be appreciated that the data so used may be interchanged under appropriate circumstances such that embodiments of the application may be practiced in sequences other than those illustrated or described herein, and that the terms "first," "second," and the like are generally used herein in a generic sense and do not limit the number of terms, e.g., the first term can be one or more than one. In addition, "and/or" in the specification and claims means at least one of connected objects, a character "/", and generally means that the former and latter related objects are in an "or" relationship.
The following describes in detail a model training method, a material recall method, and a device provided in the embodiments of the present application with reference to the accompanying drawings through specific embodiments and application scenarios thereof.
The embodiment of the application provides a model training method, and fig. 1 is a schematic flow diagram of the model training method disclosed in the embodiment of the application. As shown in fig. 1, the method includes the following steps.
S110, obtaining a plurality of user feature sets corresponding to a target user within a first preset time and a material set consisting of materials clicked by the user within the first preset time, wherein the user feature sets are sets of user features corresponding to the target user each time the material is clicked.
For example, the first preset time may be 10 days, 20 days, 30 days, etc., which is not specifically limited in this application.
It should be noted that, each time a user clicks on a material, the user may correspond to multiple user features, and the multiple user features corresponding to the user constitute a user feature set corresponding to the user. The user feature sets corresponding to the user when clicking the materials may be different, for example, the user feature set corresponding to the user when clicking the first material includes a tag of geography, the user feature set corresponding to the user when clicking the second material includes a tag of entertainment, and the user feature set corresponding to the user when clicking the third material includes a tag of sports.
In this application, the material set may include materials clicked by all users within a first preset time, for example, the first preset time is 10 days, all users are 10 users, and each user clicks 10 materials every day, so that the material set includes 1000 materials.
And S120, determining a feature vector corresponding to the user feature set based on the user feature set.
S130, performing iterative training on a target model to be trained through a plurality of feature vectors, wherein the feature vectors are input into the target model to be trained in each iterative training, and an interest preference vector corresponding to the target user and output by the target model to be trained is obtained.
After a plurality of feature vectors corresponding to the target user are obtained, the feature vectors are input into a target model to be trained for iterative training, and the trained target model can output an interest preference vector corresponding to the user according to the input feature vectors corresponding to the user.
In this application, the target model may be a three-layer Deep Neural Network (DNN) model, the numbers of neurons in three layers of the DNN model may be 1024, 512, and 256, respectively, and the neurons in the third layer output interest preference vectors corresponding to the user.
In one implementation, the target model to be trained may be trained using an Adam optimizer.
S140, determining the estimated click probability of the target material clicked by the target user in the first preset time based on the interest preference vector and a material vector matrix constructed according to the material set.
It should be noted that the dimension of the material vector matrix constructed according to the material set is the same as the number of neurons in the third layer of the DNN model, and for example, in the case that the number of neurons in the third layer of the DNN model is 256, the material vector matrix V constructed according to the material set may be a vector matrix of | V | 256, where | V | represents the number of materials included in the material set, and one material corresponds to one row in the material vector matrix V.
S150, under the condition that the estimated click probability reaches a target value, a trained target model is obtained.
The learning task of the target model requires that the estimated click probability of the target material clicked by the target user within the first preset time is determined to be the maximum based on the interest preference vector output by the target model and corresponding to the target user and the material vector matrix constructed according to the material set, and the estimated click probability of other materials not clicked by the target user within the first preset time is determined to be smaller. Based on the method, the trained target model is obtained under the condition that the estimated click probability of the target material clicked by the target user in the first preset time reaches the maximum value.
The embodiment of the application provides a model training method, which includes the steps of obtaining a plurality of user feature sets corresponding to a target user within first preset time, obtaining a material set consisting of materials clicked by the user within the first preset time, determining feature vectors corresponding to the user feature sets based on the user feature sets, conducting iterative training on a target model to be trained through the feature vectors, then determining click probability of the target materials clicked by a target user within the first preset time based on interest preference vectors output by the target model to be trained and corresponding to the target user and a material vector matrix constructed according to the material set, obtaining a trained target model under the condition that the estimated click probability reaches a target value, outputting the interest preference vectors corresponding to the user based on the user feature sets of the trained target model, and indicating high accuracy of interest preference of the user.
In this embodiment of the application, the determining an estimated click probability of the target material clicked by the target user within the first preset time based on the interest preference vector and a material vector matrix constructed according to the material set may include: and performing vector inner product operation on the interest preference vector and a target material vector, performing normalization processing, and determining the estimated click probability of the target material clicked by the target user in the first preset time, wherein the target material vector is a vector corresponding to the target material in a material vector matrix constructed according to the material set.
In an implementation manner, the performing a vector inner product operation on the interest preference vector and a target material vector, performing a normalization process, and determining an estimated click probability of a target material clicked by the target user within the first preset time may include: determining the estimated click probability of the target material clicked by the target user within the first preset time through the following formula;
wherein, P (v) i | u) is the estimated click probability of the target material clicked by the target user within the first preset time, v i Is the target material vector, u is the interest preference vector, V is the material vector matrix, V is j Is any material vector in the material vector matrix.
In this application, the determining a feature vector corresponding to the user feature set based on the user feature set may include: respectively encoding a plurality of user features in the user feature set into vectors; and splicing a plurality of vectors to obtain a feature vector corresponding to the user feature set. For example, a plurality of user features in the user feature set may be embedded into embedding in a low-dimensional manner, and concat to obtain a feature vector corresponding to the user feature set, and then the obtained feature vector is input into the target model.
In the embodiment of the present application, the user characteristics in the user characteristic set may include at least one of basic information of the user, an interest picture of the user, and behavior characteristics of the user. In the present application, the basic information of the user may include at least one of a gender of the user, an age of the user, a model of the user using the electronic device, and an area in which the user is located. The interest representation of the user may include a user's multi-level interest tags, which may include a primary interest tag, a secondary interest tag, and a tertiary interest tag, illustratively, the primary interest tag may be sports, the secondary interest tag may be basketball, the tertiary interest tag may be basketball player A, or the primary interest tag may be entertainment, the secondary interest tag may be an entertainment star in area A, the tertiary interest tag may be an entertainment star in area A, etc. The behavioral characteristics of the user may include, but are not limited to, attention bloggers, search terms, and the like.
An embodiment of the present application further provides a material recall method, and fig. 2 is a schematic flow chart of the material recall method disclosed in the embodiment of the present application. As shown in fig. 2, the method includes the following steps.
S210, user feature sets corresponding to a plurality of users are respectively obtained.
Illustratively, the user feature set may include gender, age, model of electronic device used, area of interest, primary interest tags, secondary interest tags, tertiary interest tags, bloggers of interest, search terms, and the like of the user.
S220, determining a feature vector corresponding to the user feature set based on the user feature set.
In the application, a plurality of user features in the user feature set can be respectively encoded into vectors, and the feature vectors corresponding to the user feature set are obtained by splicing the plurality of vectors. Illustratively, a plurality of user features in the user feature set may be respectively embedded into embedding in a low-dimensional manner, and concat is spliced together to obtain a feature vector corresponding to the user feature set.
And S230, respectively inputting each feature vector in the plurality of feature vectors into a target model, and acquiring a plurality of interest preference vectors corresponding to the user and output by the target model.
It should be noted that the target model is obtained by training according to the training method of the model described above.
S240, clustering a plurality of interest preference vectors to determine at least one user group.
It should be noted that, if the interest preference vectors of two users are similar, the interest preferences of the two users are similar, and the users in each user group are users with similar interest preferences.
In the method, a plurality of interest preference vectors are clustered, then users with similar interest preference vectors belong to the same user group, users with larger difference of the interest preference vectors belong to different user groups, and each user belongs to one user group.
In one implementation, clustering may be performed by a clustering algorithm, kmeans.
In addition, the heat degree and the personalization degree of the hot spot materials in the user groups can be balanced by controlling the number of the user groups, under the condition that the number of the user groups is small, the number of the users in each user group is relatively large, the higher the heat degree of the hot spot materials in the user group is, the lower the personalization degree is, under the condition that the number of the users in each user group is large, the lower the heat degree of the hot spot materials in the user group is, and the higher the personalization degree is.
And S250, respectively acquiring the click times of the users in the target user group on the plurality of materials within second preset time, wherein the target user group is any user group in the at least one user group.
In the application, the second preset time may be the current time, that is, the number of times that the user in the target user group clicks on the plurality of materials may be obtained by counting the condition that the user group clicks on the pushed materials in real time.
In addition, the second preset time may also be a time period, which is not specifically limited in this application.
And S260, determining a hot spot material list corresponding to the target user group based on the click times.
In the application, the hot spot material list corresponding to the target user group can be determined through reverse-narrative arrangement of the click times of the users in the target user group on a plurality of materials.
For example, if the users in the target user group click on the material a 300 times, click on the material B500 times, click on the material C600 times, click on the material D200 times, click on the material E20 times, and click on the material F5 times, the hot material list corresponding to the target user group is as follows:
| material(s) | Number of clicks |
| Material C | 600 times (one time) |
| |
500 times (times) |
| |
300 times (twice) |
| Material D | 200 times (one time) |
| Material E | 20 times (twice) |
| Material F | 5 times (twice) |
S270, recalling materials for users in the target user group based on the hot spot material list.
When a push is issued for a certain user in the target user group, the unread materials in the top N positions can be taken from the hot spot material list corresponding to the target user group in real time as a recall material set, and for the user to recall the materials, it is required to say that N is an integer greater than 0.
The embodiment of the application provides a material recalling method, which includes the steps of respectively obtaining user feature sets corresponding to users, determining feature vectors corresponding to the user feature sets based on the user feature sets, respectively inputting each feature vector in a plurality of feature vectors into a target model, obtaining a plurality of interest preference vectors corresponding to the users and output by the target model, then determining at least one user group by clustering the interest preference vectors, respectively obtaining the number of times of clicking on the plurality of materials by the users of the target user group within a second preset time, determining a hot point material list corresponding to the target user group based on the number of times of clicking on the plurality of materials by the users of the target user group, and recalling the materials for the users in the target user group based on the hot point material list. According to the material recalling method, the relevance between the materials recalled by the users and the interest preference of the users is high, the users are clustered into at least one user group, and then the materials are recalled for the users in the user group based on the interest preference of each user group, so that the calculated amount of online service can be reduced.
According to the training method of the model provided by the embodiment of the application, the execution subject can be a training device of the model. In the embodiment of the present application, a method for executing model training by using a model training apparatus is taken as an example, and the model training apparatus provided in the embodiment of the present application is described.
Fig. 3 is a schematic structural diagram of a training apparatus for a model according to an embodiment of the present application. As shown in fig. 3, the training apparatus 300 for a model includes: an acquisition module 310, a first determination module 320, a training module 330, a second determination module 340, and a get module 350.
In the present application, the obtaining module 310 is configured to obtain a plurality of user feature sets corresponding to a target user within a first preset time, and a material set composed of materials clicked by the user within the first preset time, where the user feature set is a set of user features corresponding to the target user each time the target user clicks the material; a first determining module 320, configured to determine, based on the user feature set, a feature vector corresponding to the user feature set; the training module 330 is configured to perform iterative training on a target model to be trained through a plurality of feature vectors, where in each iterative training, the feature vectors are input into the target model to be trained, and an interest preference vector corresponding to the target user and output by the target model to be trained is obtained; a second determining module 340, configured to determine, based on the interest preference vector and a material vector matrix constructed according to the material set, an estimated click probability of a target material clicked by the target user within the first preset time; an obtaining module 350, configured to obtain a trained target model when the estimated click probability reaches a target value.
In one implementation, the determining, by the second determining module 340, an estimated click probability of the target material clicked by the target user within the first preset time based on the interest preference vector and a material vector matrix constructed according to the material set includes: and performing vector inner product operation on the interest preference vector and a target material vector, performing normalization processing, and determining the estimated click probability of the target material clicked by the target user in the first preset time, wherein the target material vector is a vector corresponding to the target material in a material vector matrix constructed according to the material set.
In one implementation, the performing, by the second determining module 340, vector inner product operation on the interest preference vector and a target material vector, performing normalization processing, and determining an estimated click probability of a target material clicked by the target user within the first preset time includes: determining the estimated click probability of the target material clicked by the target user within the first preset time through the following formula;
wherein, P (v) i | u) is the estimated click probability of the target material clicked by the target user in the first preset time, v i Is the target material vector, u is the interest preference vector, V is the material vector matrix, V is j Is any material vector in the material vector matrix.
In one implementation, the determining module 320 determines a feature vector corresponding to the user feature set based on the user feature set, including: respectively encoding a plurality of user features in the user feature set into vectors; and splicing a plurality of vectors to obtain a feature vector corresponding to the user feature set.
In one implementation, the user characteristics in the user characteristic set include at least one of basic information of the user, interest pictures of the user, and behavior characteristics of the user.
The training device of the model in the embodiment of the present application may be an electronic device, or may be a component in an electronic device, such as an integrated circuit or a chip. The electronic device may be a terminal, or may be a device other than a terminal.
The training device for the model provided by the embodiment of the application can realize each process realized by the embodiment of the model training method, and is not repeated here to avoid repetition.
According to the material recalling method provided by the embodiment of the application, the execution main body can be a material recalling device. In the embodiment of the present application, a method for executing a recall of a material by a recall device of a material is taken as an example, and the recall device of a material provided in the embodiment of the present application is described.
Fig. 4 is a schematic structural diagram of a material recall device according to an embodiment of the present application. As shown in fig. 4, the material recalling apparatus 400 includes: a first acquisition module 410, a first determination module 420, a second acquisition module 430, a second determination module 440, a third acquisition module 450, a third determination module 460, and a recall module 470.
In the present application, the first obtaining module 410 is configured to obtain user feature sets corresponding to a plurality of users respectively; a first determining module 420, configured to determine, based on the user feature set, a feature vector corresponding to the user feature set; a second obtaining module 430, configured to input each of the feature vectors into a target model, and obtain a plurality of interest preference vectors corresponding to the user output by the target model; a second determining module 440, configured to determine at least one user group by clustering a plurality of the interest preference vectors; a third obtaining module 450, configured to obtain the number of clicks of multiple materials by a user in a target user group within a second preset time, where the target user group is any user group in the at least one user group; a third determining module 460, configured to determine, based on the number of clicks, a hot spot material list corresponding to the target user group; a recall module 470, configured to recall the items for the users in the target user group based on the hot spot item list; and the target model is obtained by training according to the training method of the model.
The material recalling device in the embodiment of the present application may be an electronic device, and may also be a component in the electronic device, such as an integrated circuit or a chip. The electronic device may be a terminal, or may be a device other than a terminal.
The material recall device provided by the embodiment of the application can realize each process realized by the material recall method embodiment, and is not repeated here for avoiding repetition.
Optionally, as shown in fig. 5, an electronic device 500 is further provided in this embodiment of the present application, and includes a processor 501 and a memory 502, where the memory 502 stores a program or an instruction that can be executed on the processor 501, and when the program or the instruction is executed by the processor 501, the steps of the embodiment of the model training method or the material recall method are implemented, and the same technical effect can be achieved, and are not described again to avoid repetition.
It should be noted that the electronic device in the embodiment of the present application includes the mobile electronic device and the non-mobile electronic device described above.
The embodiment of the present application further provides a readable storage medium, where a program or an instruction is stored on the readable storage medium, and when the program or the instruction is executed by a processor, the program or the instruction implements each process of the embodiment of the model training method or the material recall method, and can achieve the same technical effect, and in order to avoid repetition, details are not repeated here.
The processor is the processor in the electronic device described in the above embodiment. The readable storage medium includes a computer readable storage medium, such as a computer read only memory ROM, a random access memory RAM, a magnetic or optical disk, and the like.
The embodiment of the present application further provides a chip, the chip includes a processor and a communication interface, the communication interface is coupled to the processor, the processor is configured to run a program or an instruction, implement each process of the training of the model or the recall method embodiment of the material, and achieve the same technical effect, and in order to avoid repetition, the details are not repeated here.
It should be understood that the chips mentioned in the embodiments of the present application may also be referred to as system-on-chip, system-on-chip or system-on-chip, etc.
Embodiments of the present application provide a computer program product, which is stored in a storage medium and executed by at least one processor to implement the processes of the embodiments of the training method for the model or the material recall method for the model, and achieve the same technical effects, and therefore, the details are not repeated herein to avoid repetition.
It should be noted that, in this document, 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 phrases "comprising a component of' 8230; \8230;" does not exclude the presence of another like element in a process, method, article, or apparatus that comprises the element. Further, it should be noted that the scope of the methods and apparatus of the embodiments of the present application is not limited to performing the functions in the order illustrated or discussed, but may include performing the functions in a substantially simultaneous manner or in a reverse order based on the functions involved, e.g., the methods described may be performed in an order different than that described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.
Through the above description of the embodiments, those skilled in the art will clearly understand that the method of the above embodiments can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware, but in many cases, the former is a better implementation manner. Based on such understanding, the technical solutions of the present application may be embodied in the form of a computer software product, which is stored in a storage medium (such as ROM/RAM, magnetic disk, optical disk) and includes instructions for enabling a terminal (such as a mobile phone, a computer, a server, or a network device) to execute the method according to the embodiments of the present application.
While the present embodiments have been described with reference to the accompanying drawings, it is to be understood that the invention is not limited to the precise embodiments described above, which are meant to be illustrative and not restrictive, and that various changes may be made therein by those skilled in the art without departing from the spirit and scope of the invention as defined by the appended claims.
Claims (10)
1. A method of training a model, comprising:
the method comprises the steps of obtaining a plurality of user feature sets corresponding to a target user within first preset time and a material set consisting of materials clicked by the user within the first preset time, wherein the user feature sets are sets of user features corresponding to the target user each time the target user clicks the materials;
determining a feature vector corresponding to the user feature set based on the user feature set;
performing iterative training on a target model to be trained through a plurality of feature vectors, wherein the feature vectors are input into the target model to be trained in each iterative training, and an interest preference vector corresponding to the target user and output by the target model to be trained is obtained;
determining the estimated click probability of the target material clicked by the target user within the first preset time based on the interest preference vector and a material vector matrix constructed according to the material set;
and obtaining a trained target model under the condition that the estimated click probability reaches a target value.
2. The training method according to claim 1, wherein the determining the estimated click probability of the target material clicked by the target user within the first preset time based on the interest preference vector and a material vector matrix constructed according to the material set comprises:
and performing vector inner product operation on the interest preference vector and a target material vector, performing normalization processing, and determining the estimated click probability of the target material clicked by the target user in the first preset time, wherein the target material vector is a vector corresponding to the target material in a material vector matrix constructed according to the material set.
3. The training method according to claim 2, wherein the performing vector inner product operation on the interest preference vector and a target material vector, performing normalization processing, and determining the estimated click probability of the target material clicked by the target user within the first preset time comprises:
determining the estimated click probability of the target material clicked by the target user within the first preset time through the following formula;
4. The training method of claim 1, wherein the determining a feature vector corresponding to the user feature set based on the user feature set comprises:
respectively encoding a plurality of user features in the user feature set into vectors;
and splicing a plurality of vectors to obtain a feature vector corresponding to the user feature set.
5. The training method of claim 1, wherein the user features in the user feature set comprise at least one of basic information of the user, user interest pictures and user behavior features.
6. A method for recalling material, comprising:
respectively acquiring user feature sets corresponding to a plurality of users;
determining a feature vector corresponding to the user feature set based on the user feature set;
respectively inputting each feature vector in the feature vectors into a target model, and acquiring a plurality of interest preference vectors corresponding to the user and output by the target model;
determining at least one user group by clustering a plurality of the interest preference vectors;
respectively acquiring the number of clicks of a plurality of materials by users in a target user group within second preset time, wherein the target user group is any user group in the at least one user group;
determining a hot spot material list corresponding to the target user group based on the click times;
recalling materials for users in the target user group based on the hot material list;
wherein the target model is obtained by training according to the training method of the model of any one of the preceding claims 1 to 5.
7. An apparatus for training a model, comprising:
the system comprises an acquisition module, a display module and a display module, wherein the acquisition module is used for acquiring a plurality of user feature sets corresponding to a target user within first preset time and a material set consisting of materials clicked by the user within the first preset time, and the user feature sets are the sets of user features corresponding to the target user when clicking the materials each time;
the first determining module is used for determining a feature vector corresponding to the user feature set based on the user feature set;
the training module is used for performing iterative training on a target model to be trained through a plurality of feature vectors, wherein the feature vectors are input into the target model to be trained in each iterative training, and an interest preference vector which is output by the target model to be trained and corresponds to the target user is obtained;
the second determination module is used for determining the estimated click probability of the target material clicked by the target user within the first preset time based on the interest preference vector and a material vector matrix constructed according to the material set;
and the obtaining module is used for obtaining the trained target model under the condition that the estimated click probability reaches the target value.
8. A material recall device, comprising:
the first acquisition module is used for respectively acquiring user feature sets corresponding to a plurality of users;
the first determination module is used for determining a feature vector corresponding to the user feature set based on the user feature set;
a second obtaining module, configured to input each of the feature vectors into a target model, and obtain a plurality of interest preference vectors corresponding to the user and output by the target model;
a second determining module for determining at least one user group by clustering a plurality of the interest preference vectors;
a third obtaining module, configured to respectively obtain click times of multiple materials by users in a target user group within a second preset time, where the target user group is any user group in the at least one user group;
a third determining module, configured to determine, based on the number of clicks, a hot material list corresponding to the target user group;
a recall module, configured to recall the materials for the users in the target user group based on the hot spot material list;
wherein the target model is obtained by training according to the training method of the model of any one of the preceding claims 1 to 5.
9. An electronic device comprising a processor and a memory, said memory storing a program or instructions executable on said processor, said program or instructions, when executed by said processor, implementing the steps of the training method of the model according to any one of claims 1-5, or said program or instructions, when executed by said processor, implementing the steps of the recall method of the material according to claim 6.
10. A readable storage medium, characterized in that it stores thereon a program or instructions which, when executed by a processor, implement the steps of a training method of a model according to any one of claims 1-5, or which, when executed by said processor, implement the steps of a recall method of a material according to claim 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211604889.5A CN115730217A (en) | 2022-12-14 | 2022-12-14 | Model training method, material recalling method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211604889.5A CN115730217A (en) | 2022-12-14 | 2022-12-14 | Model training method, material recalling method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115730217A true CN115730217A (en) | 2023-03-03 |
Family
ID=85301358
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211604889.5A Pending CN115730217A (en) | 2022-12-14 | 2022-12-14 | Model training method, material recalling method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115730217A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117556136A (en) * | 2023-11-13 | 2024-02-13 | 书行科技(北京)有限公司 | Article recall method and device, electronic equipment and computer readable storage medium |
-
2022
- 2022-12-14 CN CN202211604889.5A patent/CN115730217A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117556136A (en) * | 2023-11-13 | 2024-02-13 | 书行科技(北京)有限公司 | Article recall method and device, electronic equipment and computer readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111538912B (en) | Content recommendation method, device, equipment and readable storage medium | |
| CN110378731B (en) | Method, device, server and storage medium for acquiring user portrait | |
| CN111553754B (en) | Updating method and device of behavior prediction system | |
| CN110717098B (en) | Meta-path-based context-aware user modeling method and sequence recommendation method | |
| CN110717099B (en) | Method and terminal for recommending film | |
| US10354184B1 (en) | Joint modeling of user behavior | |
| CN105022754B (en) | Object classification method and device based on social network | |
| CN110688482B (en) | Multi-label identification method, training method and device | |
| CN112434151A (en) | Patent recommendation method and device, computer equipment and storage medium | |
| CN112364204B (en) | Video searching method, device, computer equipment and storage medium | |
| CN111506820B (en) | Recommendation model, recommendation method, recommendation device, recommendation equipment and recommendation storage medium | |
| CN110008397B (en) | Recommendation model training method and device | |
| CN107786943A (en) | A kind of tenant group method and computing device | |
| CN114036406A (en) | Recommendation method and system based on graph contrast learning and social network enhancement | |
| CN108629010A (en) | It is a kind of to recommend method based on the web services of theme and service combination information | |
| CN115730217A (en) | Model training method, material recalling method and device | |
| CN113327132A (en) | Multimedia recommendation method, device, equipment and storage medium | |
| CN112765450A (en) | Recommended content determining method, recommended content determining device and storage medium | |
| CN110769288A (en) | Video cold start recommendation method and system | |
| CN114223012A (en) | Push object determination method and device, terminal equipment and storage medium | |
| CN110162714B (en) | Content pushing method, device, computing equipment and computer readable storage medium | |
| CN113516524A (en) | Method and device for pushing information | |
| CN111897910A (en) | Information pushing method and device | |
| CN117786234B (en) | Multimode resource recommendation method based on two-stage comparison learning | |
| CN116777529B (en) | Object recommendation method, device, equipment, storage medium and program product |
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 |