CN114707063A - Commodity recommendation method and device, electronic equipment and storage medium - Google Patents
Commodity recommendation method and device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN114707063A CN114707063A CN202210310761.1A CN202210310761A CN114707063A CN 114707063 A CN114707063 A CN 114707063A CN 202210310761 A CN202210310761 A CN 202210310761A CN 114707063 A CN114707063 A CN 114707063A
- Authority
- CN
- China
- Prior art keywords
- shopping
- social
- sample set
- behavior sample
- dissociation
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 62
- 230000006399 behavior Effects 0.000 claims abstract description 151
- 230000011273 social behavior Effects 0.000 claims abstract description 147
- 238000010494 dissociation reaction Methods 0.000 claims abstract description 100
- 230000005593 dissociations Effects 0.000 claims abstract description 96
- 239000013598 vector Substances 0.000 claims abstract description 47
- 238000012549 training Methods 0.000 claims abstract description 24
- 238000012512 characterization method Methods 0.000 claims description 110
- 238000012937 correction Methods 0.000 claims description 34
- 238000004590 computer program Methods 0.000 claims description 19
- 238000004364 calculation method Methods 0.000 claims description 12
- 238000012545 processing Methods 0.000 abstract description 7
- 230000008569 process Effects 0.000 description 21
- 238000010586 diagram Methods 0.000 description 10
- 230000006870 function Effects 0.000 description 4
- 239000011159 matrix material Substances 0.000 description 4
- 230000009471 action Effects 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000003796 beauty Effects 0.000 description 2
- 239000002537 cosmetic Substances 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 206010037660 Pyrexia Diseases 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 238000013144 data compression Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
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
- 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
- G06F16/9536—Search customisation based on social or collaborative filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/01—Social networking
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Human Resources & Organizations (AREA)
- Health & Medical Sciences (AREA)
- Economics (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The application provides a commodity recommendation method and device, electronic equipment and a storage medium, and belongs to the technical field of data processing. The method comprises the following steps: acquiring historical shopping behaviors and social behaviors of a current user; respectively representing the historical shopping behaviors and the social behaviors of the current user by vectors; inputting the vector representation of the historical shopping behavior and the vector representation of the social behavior of the current user into a pre-trained commodity recommendation model, and determining a target commodity recommended to the current user through the pre-trained commodity recommendation model; the pre-trained commodity recommendation model is obtained by respectively performing dissociation representation on a shopping behavior sample set and a social behavior sample set based on shopping behavior sample sets and social behavior sample sets of a plurality of historical users and training according to an influence relation between a dissociation representation result of the shopping behavior sample set and a dissociation representation result of the social behavior sample set. The commodity recommendation method and the commodity recommendation device aim to improve commodity recommendation accuracy.
Description
Technical Field
The embodiment of the application relates to the technical field of data processing, in particular to a commodity recommendation method and device, electronic equipment and a storage medium.
Background
At present, in order to improve the shopping experience of a user, a commodity recommendation system is developed rapidly, and the commodity recommendation system considers the social relationship of the user as an important factor influencing shopping selection.
However, in the existing commodity recommendation system based on the social relationship of the user, the social relationship of the user is modeled as a model influencing the shopping decision of the user, a large number of potential factors behind social information are ignored, and the factors influencing social preference and shopping preference are easily coupled with each other, so that the commodity recommendation accuracy is reduced.
Disclosure of Invention
The embodiment of the application provides a commodity recommendation method, a commodity recommendation device, electronic equipment and a storage medium, and aims to improve the accuracy of commodity recommendation.
In a first aspect, an embodiment of the present application provides a commodity recommendation method, where the method includes:
acquiring historical shopping behaviors and social behaviors of a current user;
respectively representing the historical shopping behaviors and the social behaviors of the current user by vectors;
inputting a pre-trained commodity recommendation model into the vector representation of the historical shopping behavior and the vector representation of the social behavior of the current user, and determining a target commodity recommended to the current user through the pre-trained commodity recommendation model;
the pre-trained commodity recommendation model is obtained by respectively performing dissociation representation on a shopping behavior sample set and a social behavior sample set based on shopping behavior sample sets and social behavior sample sets of a plurality of historical users and training according to an influence relation between a dissociation representation result of the shopping behavior sample set and a dissociation representation result of the social behavior sample set.
Optionally, the commodity recommendation model is obtained by training according to the following steps:
acquiring historical shopping behaviors and historical social behaviors of a plurality of historical users;
establishing the shopping behavior sample set according to the respective historical shopping behaviors of a plurality of historical users, and establishing the social behavior sample set according to the respective historical social behaviors of the plurality of historical users;
respectively inputting the shopping behavior sample set and the social behavior sample set into a preset model, wherein the preset model comprises a plurality of variational automatic encoders;
respectively determining respective dissociation representation results of the shopping behavior sample set and the social behavior sample set through a preset model;
according to the dissociation characterization result of the shopping behavior sample set, correcting the dissociation characterization result of the social behavior sample set, and updating the model parameters of the preset model according to the correction result;
according to the dissociation characterization result of the social behavior sample set, correcting the dissociation characterization result of the shopping behavior sample set, and updating the model parameters of the preset model according to the correction result;
and after the preset model is iteratively updated for preset times, the commodity recommendation model is obtained.
Optionally, after determining respective dissociation characterization results of the shopping behavior sample set and the social behavior sample set through a preset model, the method further includes:
determining real influence parameters among the plurality of historical users according to respective dissociation representation results of the shopping behavior sample set and the social behavior sample set;
correcting the dissociation characterization result of the social behavior sample set according to the dissociation characterization result of the shopping behavior sample set, wherein the correction comprises the following steps:
correcting the dissociation characterization result of the social behavior sample set according to the real influence parameter and the social correction coefficient;
correcting the dissociation characterization result of the shopping behavior sample set according to the dissociation characterization result of the social behavior sample set, wherein the correction comprises the following steps:
and correcting the dissociation characterization result of the shopping behavior sample set according to the real influence parameter and the shopping correction coefficient.
Optionally, the dissociated representation result of the shopping behavior sample set comprises a plurality of first channels, each of the first channels represents a shopping intention of the user; the dissociation characterization result of the social behavior sample set comprises a plurality of second channels, and each second channel characterizes one social intention of the user;
determining real influence parameters among the plurality of historical users according to the respective dissociation characterization results of the shopping behavior sample set and the social behavior sample set, wherein the determination comprises the following steps:
respectively calculating the shopping representation similarity of the users of the plurality of first channels;
determining average shopping characterization similarity among a plurality of historical users according to the shopping characterization similarity of each user of the plurality of first channels;
respectively calculating the user social representation similarity of each of the plurality of second channels;
determining average social representation similarity among a plurality of historical users according to the respective user social representation similarity of the plurality of second channels;
and determining the real influence parameters according to the average shopping characterization similarity and the average social characterization similarity.
Optionally, the calculation formula of the shopping characterization similarity of the users in each of the plurality of first channels is as follows:
the calculation formula of the average shopping characterization similarity is as follows:
in the two formulas, X is the historical shopping behavior of the user;
is the k-thXThe shopping representation similarity of the users of the first channel; simXCharacterizing similarity for average purchases;
characterization for the k-th after dissociationXThe dissociation characterization results corresponding to the first channels; k is not less than 1X≤KX，KXIs the total number of channels of the first channel.
Optionally, the calculation formula of the user social representation similarity of each of the plurality of second channels is:
the calculation formula of the average social representation similarity is as follows:
in the two formulas, A is the historical social behavior of the user;
is the k-thAThe social representation similarity of the users of the second channel; simACharacterize the similarity for average sociality;
characterization for the k-th after dissociationAThe dissociation characterization results corresponding to the second channels; k is not less than 1A≤KA，KAThe total number of channels of the second channel.
Optionally, the calculation formula of the real impact parameter is:
sim＝simX·simA
in the formula, simXCharacterizing similarity for average purchases; simACharacterize the similarity for average sociality;
according to the real influence parameters and the shopping correction coefficients, a formula for correcting the dissociation characterization results of the shopping behavior sample set is as follows:
in the formula, alphaXCorrecting the coefficient for shopping;
according to the real influence parameters and the social correction coefficient, a formula for correcting the dissociation characterization results of the social behavior sample set is as follows:
in the formula, alphaAIs a social correction factor.
In a second aspect, an embodiment of the present application provides an article recommendation device, where the device includes:
the acquisition module is used for acquiring historical shopping behaviors and social behaviors of a current user;
the vector representation module is used for respectively representing the historical shopping behaviors and the social behaviors of the current user by vectors;
the commodity recommendation module is used for inputting the vector representation of the historical shopping behavior and the vector representation of the social behavior of the current user into a pre-trained commodity recommendation model, and determining a target commodity recommended to the current user through the pre-trained commodity recommendation model; the pre-trained commodity recommendation model is obtained by respectively performing dissociation representation on a shopping behavior sample set and a social behavior sample set based on shopping behavior sample sets and social behavior sample sets of a plurality of historical users and training according to an influence relation between a dissociation representation result of the shopping behavior sample set and a dissociation representation result of the social behavior sample set.
In a third aspect, an embodiment of the present application provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, where the processor executes the computer program to implement the merchandise recommendation method according to the first aspect of the embodiment.
In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the method for recommending an article according to the first aspect of the embodiment is implemented.
Has the advantages that:
the method comprises the steps of obtaining historical shopping behaviors and social behaviors of a current user, representing the historical shopping behaviors and the social behaviors of the current user by vectors, inputting the vector representations of the historical shopping behaviors and the social behaviors of the current user into a pre-trained commodity recommendation model, and determining target commodities recommended to the current user through the pre-trained commodity recommendation model.
In the method, a commodity recommendation model is used for recommending target commodities to users according to historical shopping behaviors and social behaviors of current users, the commodity recommendation model is trained on a shopping behavior sample set and a social behavior sample set of a plurality of historical users, in the training process, the shopping behavior sample set and the social behavior sample set are subjected to dissociation characterization, so that the shopping behavior and the social behavior are decomposed into finer granularity, then according to the influence relationship between the dissociation representation result of the shopping behavior sample set and the dissociation representation result of the social behavior sample set, the shopping behavior and the social behavior are interacted, the commodity recommendation model obtained through training determines the influence of the social behaviors of the user on the shopping behaviors on a finer granularity level, and can provide more suitable commodities for the current user, so that the commodity recommendation accuracy is improved.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the description of the embodiments of the present application will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art that other drawings can be obtained according to these drawings without inventive exercise.
Fig. 1 is a flowchart illustrating steps of a method for recommending merchandise according to an embodiment of the present application;
FIG. 2 is a flowchart illustrating steps for training a recommendation model according to an embodiment of the present application;
fig. 3 is a functional block diagram of a product recommendation device according to an embodiment of the present application.
Detailed Description
The technical solutions in the embodiments of the present application will be clearly and completely described 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, 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 application.
The shopping behavior and the social behavior of the user are usually embodied by the interests and hobbies of the user, so that the shopping behavior and the social behavior of the user are connected and influenced mutually, for example, if the user likes to purchase electronic equipment, the user usually pays attention to bloggers in the scientific and electronic field, or if the user likes to make up, the user pays attention to some bloggers in the making up; similarly, the user pays attention to the blogger in the science and technology or electronic field, and may influence the user to purchase the electronic equipment, or the user pays attention to the blogger of beauty cosmetics, and the user is more likely to purchase the products of beauty cosmetics.
According to the method and the device, the shopping behavior and the social behavior of the user are dissociated, and the interaction relation among all dissociation results between the shopping behavior and the social behavior is excavated at a finer-grained level, so that the accuracy in recommending commodities to the user is improved.
Referring to fig. 1, a flowchart illustrating steps of a product recommendation method in an embodiment of the present invention is shown, and as shown in fig. 1, the method may specifically include the following steps:
s101: and acquiring historical shopping behaviors and social behaviors of the current user.
In an actual implementation process, all or part of historical shopping behaviors and social behaviors of a current user can be acquired, and if the current user wants to shop on a shopping client, the historical shopping behaviors of the user on the shopping client can be acquired, for example, by acquiring a shopping record of the current user in a historical order list and acquiring the social behaviors of the current user on the shopping client, such as a shop concerned by the current user, a shop frequently visited, a favorite commodity, other shopping users concerned or a blogger; under the condition of obtaining the permission or the permission of the current user, historical shopping behaviors and social behaviors of the user on other clients can be obtained.
S102: and respectively representing the historical shopping behaviors and the social behaviors of the current user by vectors.
After the historical shopping behaviors and the social behaviors of the current user are obtained, the historical shopping behaviors and the social behaviors are respectively expressed by vectors so as to meet the requirements of input conditions of a commodity recommendation model. Illustratively, for any current user, the shopping behavior is a vector consisting of 0 and 1, which respectively represent that no goods are purchased/purchased; similarly, the social behavior is a vector consisting of 0 and 1, which indicates that the user is not paying attention to other users, stores or bloggers.
S103: inputting the vector representation of the historical shopping behavior and the vector representation of the social behavior of the current user into a pre-trained commodity recommendation model, and determining the target commodity recommended to the current user through the pre-trained commodity recommendation model.
The pre-trained commodity recommendation model is obtained by respectively performing dissociation characterization on a shopping behavior sample set and a social behavior sample set based on shopping behavior sample sets and social behavior sample sets of a plurality of historical users and training according to an influence relationship between dissociation characterization results of the shopping behavior sample set and dissociation characterization results of the social behavior sample set.
The commodity recommendation model carries out dissociation representation on the shopping behavior sample set and the social behavior sample set in the training process, further decomposes the preference of a user when the user executes any intention in the shopping behavior and social behavior processes into finer granularity, and then enables the shopping behavior and the social behavior to interact according to the influence relationship between the dissociation representation result of the shopping behavior sample set and the dissociation representation result of the social behavior sample set.
Referring to fig. 2, a flowchart illustrating steps of training a product recommendation model in an embodiment of the present invention is shown, and as shown in fig. 2, the product recommendation model is obtained by training according to the following steps:
s201: and acquiring historical shopping behaviors and historical social behaviors of a plurality of historical users.
The number of history users is not limited in the embodiment, and the user-defined setting can be carried out according to the actual requirement; the historical shopping behavior of the user can be based on the historical shopping records of the same shopping client or different shopping clients, and the historical social behavior of the user can be based on the behavior of the same social platform or different social platforms.
S202: and constructing the shopping behavior sample set according to the respective historical shopping behaviors of the plurality of historical users, and constructing the social behavior sample set according to the respective historical social behaviors of the plurality of historical users.
In practical implementation, the shopping behavior sample set and the social behavior sample set may be matrices, and illustratively, the shopping behavior sample set is an N × M matrix X, where N rows represent N users, M columns represent M commodities, and a value X in the matrixijIs 0 or 1, xijWhen 0, it means that a certain user i has not purchased a certain commodity j, xijA certain item j is purchased for 1 on behalf of a certain user i. The social behavior sample set is a matrix A with the size of N × N, wherein N rows and N columns all represent N users, and the value a in the matrixijIs a number of 0 or 1, and,aijwhen 0, it means that a certain user i does not pay attention to another user j, aijA value of 1 represents that a certain user i is interested in another user j.
S203: and respectively inputting the shopping behavior sample set and the social behavior sample set into a preset model, wherein the preset model comprises a plurality of variation automatic encoders.
A Variational Auto-Encoder (VAE) is a structure composed of an Encoder and a decoder, wherein the Encoder is mainly used for performing dimension reduction or data compression on input initial data, namely, a process of reducing the quantity of features describing the data, the dimension reduction is performed by retaining or extracting the features from the features of the initial data, and the decoder is used for performing an inverse process thereof, namely, reconstructing based on the retained or extracted features.
In the method, an encoder is used for encoding high-dimensional effective information sparse data of a shopping behavior sample set and a social behavior sample set into low-dimensional effective information dense data, namely dissociative characterization is carried out on the shopping behavior sample set and the social behavior sample set, the dissociative characterization process is carried out in an unsupervised mode, supervision of any other additional information is not needed, the preset model is trained only by the two matrixes X and A, and the two initial matrixes are restored after dissociative characterization by the preset model.
S204: and respectively determining respective dissociation characterization results of the shopping behavior sample set and the social behavior sample set through a preset model.
In the method, a preset model comprises a plurality of encoders, a dissociation vector encoded by the encoder of each VAE is regarded as a channel, and a shopping behavior sample set and a social behavior sample set can be represented as dissociation representations of the plurality of channels, wherein the dissociation representation result of the shopping behavior sample set comprises a plurality of first channels, and each first channel represents a shopping intention of a user; the dissociated characterization result of the social behavior sample set includes a plurality of second channels, each of which characterizes a social intent of the user.
The channels represent the high-level intentions of the user, and the different dimensions of the dissociation vector represent the low-level preferences, for example, the high-level intentions represented by the channels corresponding to the shopping behaviors are the types of articles purchased by the user in the shopping process, for example, the channel 1 corresponding to the shopping behavior is clothing, the channel 2 is sports equipment, the channel 3 is an electronic product, and the low-level preferences mainly refer to the characteristics of the articles, such as the style, size, color and the like of the clothing; the channel corresponding to the social behavior represents that the high-level intention is the group attribute of the user, such as a sports enthusiast, a makeup blogger, an electronic fever friend and the like, and the low-level preference refers to the personal attribute of the user, such as age, gender, interest and the like, because a certain user may be concerned by another user due to the group attribute or the personal attribute of the user.
When the type of the goods or the attributes of the user are determined, the goods or the users which are closer to a certain type of prototype are regarded as the type in a 'prototype' mode, for example, the characterization vectors of coats, sweaters, shirts, T-shirts and the like are different, the vector corresponding to the 'middle point' of the plurality of characterization vectors is obtained by averaging the characterization vectors of the clothes, and the characterization vector corresponding to the average value is regarded as the characterization vector of the goods; and when the new commodity data enters the model, judging the distance between the characterization vector of the new commodity and the characterization vector of the 'clothes', and if the distance is smaller than a preset value, representing that the new commodity belongs to the category of the 'clothes'.
In real life, shopping behaviors and social behaviors of users are influenced mutually, for example, a user 1 purchases a commodity 1, a user 2 who also purchases the commodity 1 may be acquainted, and after the user 1 and the user 2 establish a social relationship, evaluation or use of the commodity 2 by the user 2 may influence the user 1 to purchase the commodity 2. Therefore, after obtaining the dissociation characterization result of the shopping behavior sample set and the dissociation characterization result of the social behavior sample set, the method further corrects the dissociation characterization result of the shopping behavior sample set and the dissociation characterization result of the social behavior sample set to perform an alternating iterative update.
Before correcting the dissociation characterization results of the shopping behavior sample set and the social behavior sample set, determining real influence parameters among the plurality of historical users according to the respective dissociation characterization results of the shopping behavior sample set and the social behavior sample set, which includes the following specific steps:
the shopping behavior sample set is marked as X, the social behavior sample set is marked as A, and the dissociation characterization result of the shopping behavior sample set is marked as X
The result of the dissociative characterization of the social behavior sample set is recorded as
Wherein, k is more than or equal to 1X≤KX，1≤kA≤KA，KXThe total number of channels of the first channel; kAThe total number of channels of the second channel.
A1: respectively calculating the respective user shopping representation similarity of the plurality of first channels, wherein the user shopping representation similarity represents that different users are from the kthXThe individual intent to purchase a similar degree of the item.
KthXThe calculation formula of the shopping characterization similarity of the users in the first channel is as follows:
and sequentially calculating the shopping representation similarity of the users of all the first channels.
A2: and determining the average shopping characterization similarity among a plurality of historical users according to the shopping characterization similarity of the users in the first channels.
The calculation formula of the average shopping characterization similarity is as follows:
a3: and respectively calculating the user social representation similarity of each of the plurality of second channels, wherein the user social representation similarity represents the similarity of the group attributes of the user and the user.
KthAThe calculation formula of the user social representation similarity of the second channel is as follows:
and sequentially calculating the social representation similarity of the users of all the second channels.
A4: and determining the average social representation similarity among a plurality of historical users according to the respective user social representation similarity of the plurality of second channels.
The calculation formula of the average social representation similarity is as follows:
a5: and determining the real influence parameters according to the average shopping characterization similarity and the average social characterization similarity.
Average shopping characterization similarity simXSimilarity to average social representation simAThe method embodies the closeness degree of the connection between the users in the aspects of shopping and social contact, and represents the similarity sim to the average shoppingXSimilarity to average social representation simAObtaining the true influence parameter sim by intersection, i.e. sim ═ simX·simAThe real impact parameter sim is used to reflect the interaction between users, e.g. if simXThe numerical value is larger, which indicates that the users have similarity in shopping behaviors; if simAThe larger value indicates that the users have social behavior similarity. If the product of the two is large, namely the value of the real influence parameter sim is large, the similarity of the users in both shopping and behavior is shown, but the similarity is not only in one aspect.
S205: and correcting the dissociation characterization result of the social behavior sample set according to the dissociation characterization result of the shopping behavior sample set, and updating the model parameters of the preset model according to the correction result.
In a feasible implementation manner, after a real influence parameter is obtained through calculation, the dissociation characterization result of the shopping behavior sample set is corrected according to the real influence parameter and a shopping correction coefficient, and the formula is as follows:
in the formula, alphaXCorrection factor for shopping, αXThe value of (a) can be set in a self-defined way, the greater the influence of social behaviors on shopping behaviors, alphaXThe larger.
S206: and correcting the dissociation characterization result of the shopping behavior sample set according to the dissociation characterization result of the social behavior sample set, and updating the model parameters of the preset model according to the correction result.
In a possible embodiment, the dissociation characterization result of the social behavior sample set is modified according to the real impact parameter and a social modification coefficient, and a formula is as follows:
in the formula, alphaAAs social correction factor, αAThe value of (a) can be set in a self-defined way, the greater the influence of the shopping behavior on the social behavior, alphaAThe larger.
S207: and after the preset model is iteratively updated for preset times, the commodity recommendation model is obtained.
The process of the preset model training comprises a dissociation process and a mutual correction process. For example, the preset model may first dissociate the shopping behavior sample set, and after dissociative characterization results of shopping habits of different users are obtained preliminarily, in a process of dissociating the social behavior sample set by the preset model, the preset model obtains an average shopping characterization similarity of the users according to the time of dissociating the shopping behavior sample set, so as to correct the dissociative characterization of the social behavior sample set, and the above process is regarded as one iteration in a second iteration, and the dissociative characterization result of the social behavior sample set may also correct the dissociative characterization of the shopping behavior sample set. And after the preset model is iteratively updated for preset times, the commodity recommendation model is obtained.
In a possible implementation manner, when the preset model is trained for a preset number of times, each iterative training may be based on different shopping behavior sample sets and social behavior sample sets, and each training includes a dissociation process of shopping behavior and social behavior, a modified shopping process, and a reconstruction process. Illustratively, the first training is based on the dissociative characterization of the shopping behavior sample 1 and the social behavior sample 1, the dissociative characterization of the social behavior sample is corrected by using the average shopping characterization similarity of the shopping behavior sample 1, the second training is based on the dissociative characterization of the shopping behavior sample 2 and the social behavior sample 2, the dissociative characterization of the shopping behavior sample 2 is corrected by using the average social characterization similarity of the social behavior sample 2, and the commodity recommendation model is obtained by performing alternate iterative updating when the iteration is preset.
Generally speaking, the training process of the commodity recommendation model is to dissociate the shopping behavior sample set and the social behavior sample set, solve the similarity of the dissociated characterization results of the shopping behavior sample set and the social behavior sample set, and provide smaller weight to the opposite party as correction.
After the training of the commodity recommendation model is completed, when the recommendation is carried out on the current user, the historical shopping behavior of the current user is represented by 0/1 vectors, and the historical social behavior of the current user is represented by 0/1 vectors. Taking historical shopping behavior as an example, assuming that there are A, B, C, D, E items in the item set and the current user purchased the item B, the vector of the input shopping information is [0,1,0,0,0 ]. The user has not purchased another four items A, C, D, E.
Inputting a pre-trained commodity recommendation model according to vector representation of historical shopping behaviors and vector representation of social behaviors of a current user, dissociating shopping and social preferences of the current user by the commodity recommendation model, and restoring possible shopping demands of the user through a decoder according to the dissociated low-dimensional dense vector which contains the intention and the preferences of the user. Assuming that the restored vectors are [0.1,0.9,0.3,0.4,0.7], the vectors respectively correspond to the possibility that the current user purchases five commodities, wherein the possibility of B is the maximum, since this is the known situation of user purchase, after ignoring the possibility corresponding to the purchased commodities, the commodities with the possibility greater than the preset possibility threshold are recommended to the current user, and the preset feasibility threshold and the number of recommended commodities can be set by the user according to actual application.
Referring to fig. 3, there is shown a functional block diagram of an article recommendation apparatus in an embodiment of the present invention, the apparatus including:
an obtaining module 100, configured to obtain historical shopping behaviors and social behaviors of a current user;
a vector representation module 200, configured to represent the historical shopping behavior and the social behavior of the current user by vectors respectively;
the commodity recommendation module 300 is configured to input a vector representation of a historical shopping behavior and a vector representation of a social behavior of the current user into a pre-trained commodity recommendation model, and determine a target commodity recommended to the current user according to the pre-trained commodity recommendation model; the pre-trained commodity recommendation model is obtained by respectively performing dissociation representation on a shopping behavior sample set and a social behavior sample set based on shopping behavior sample sets and social behavior sample sets of a plurality of historical users and training according to an influence relation between a dissociation representation result of the shopping behavior sample set and a dissociation representation result of the social behavior sample set.
Optionally, the apparatus further includes a model training module, specifically, the model training module includes:
the system comprises an acquisition unit, a storage unit and a processing unit, wherein the acquisition unit is used for acquiring historical shopping behaviors and historical social behaviors of a plurality of historical users;
the sample determining unit is used for constructing the shopping behavior sample set according to the respective historical shopping behaviors of a plurality of historical users and constructing the social behavior sample set according to the respective historical social behaviors of the plurality of historical users;
the input unit is used for respectively inputting the shopping behavior sample set and the social behavior sample set into a preset model, wherein the preset model comprises a plurality of variational automatic encoders;
the dissociation unit is used for respectively determining respective dissociation representation results of the shopping behavior sample set and the social behavior sample set through a preset model;
the first correction unit is used for correcting the dissociation characterization result of the social behavior sample set according to the dissociation characterization result of the shopping behavior sample set and updating the model parameters of the preset model according to the correction result;
the second correction unit is used for correcting the dissociation characterization result of the shopping behavior sample set according to the dissociation characterization result of the social behavior sample set and updating the model parameters of the preset model according to the correction result;
and the model determining unit is used for obtaining the commodity recommendation model after the preset model is iteratively updated for preset times.
Optionally, the model training module further comprises:
a real influence parameter determination unit, configured to determine a real influence parameter between the multiple historical users according to respective dissociation characterization results of the shopping behavior sample set and the social behavior sample set;
the first correction unit includes:
the first correction subunit is used for correcting the dissociation representation result of the social behavior sample set according to the real influence parameter and the social correction coefficient;
the second correcting unit includes:
and the second correction subunit corrects the dissociation characterization result of the shopping behavior sample set according to the real influence parameter and the shopping correction coefficient.
Optionally, the dissociated representation result of the shopping behavior sample set comprises a plurality of first channels, each of the first channels represents a shopping intention of the user; the dissociation characterization result of the social behavior sample set comprises a plurality of second channels, and each second channel characterizes one social intention of the user;
the real influence parameter determination unit includes:
the first similarity determining subunit is used for respectively calculating the shopping characterization similarity of the users of the plurality of first channels;
the first average subunit is used for determining the average shopping characterization similarity among a plurality of historical users according to the shopping characterization similarity of each user in the plurality of first channels;
the second similarity determining subunit is used for respectively calculating the user social representation similarity of each of the plurality of second channels;
the second average subunit is used for determining average social representation similarity among a plurality of historical users according to the respective user social representation similarity of the plurality of second channels;
and the determining subunit is used for determining the real influence parameters according to the average shopping characterization similarity and the average social characterization similarity.
The embodiment of the present application further provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and capable of running on the processor, and when the processor executes the computer program, the article recommendation method according to the embodiment is implemented.
An embodiment of the present application further provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the method for recommending an article according to the embodiment is implemented.
The embodiments in the present specification are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other.
As will be appreciated by one of skill in the art, embodiments of the present application may be provided as a method, apparatus, or computer program product. Accordingly, embodiments of the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, embodiments of the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
Embodiments of the present application are described with reference to flowchart illustrations and/or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing terminal to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing terminal to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing terminal to cause a series of operational steps to be performed on the computer or other programmable terminal to produce a computer implemented process such that the instructions which execute on the computer or other programmable terminal provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
While preferred embodiments of the present application have been described, additional variations and modifications of these embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. Therefore, it is intended that the appended claims be interpreted as including the preferred embodiment and all such alterations and modifications as fall within the true scope of the embodiments of the application.
Finally, it should also be noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or terminal 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 terminal. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, method, article, or terminal that comprises the element.
The principle and the implementation of the present application are explained herein by applying specific examples, and the above description of the embodiments is only used to help understand the method and the core idea of the present application; meanwhile, for a person skilled in the art, according to the idea of the present application, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present application.
Claims (10)
1. A method for recommending an article, the method comprising:
acquiring historical shopping behaviors and social behaviors of a current user;
respectively representing the historical shopping behaviors and the social behaviors of the current user by vectors;
inputting a pre-trained commodity recommendation model into the vector representation of the historical shopping behavior and the vector representation of the social behavior of the current user, and determining a target commodity recommended to the current user through the pre-trained commodity recommendation model;
the pre-trained commodity recommendation model is obtained by respectively performing dissociation representation on a shopping behavior sample set and a social behavior sample set based on shopping behavior sample sets and social behavior sample sets of a plurality of historical users and training according to an influence relation between a dissociation representation result of the shopping behavior sample set and a dissociation representation result of the social behavior sample set.
2. The method of claim 1, wherein the commodity recommendation model is trained according to the following steps:
acquiring historical shopping behaviors and historical social behaviors of a plurality of historical users;
establishing the shopping behavior sample set according to the respective historical shopping behaviors of a plurality of historical users, and establishing the social behavior sample set according to the respective historical social behaviors of the plurality of historical users;
respectively inputting the shopping behavior sample set and the social behavior sample set into a preset model, wherein the preset model comprises a plurality of variational automatic encoders;
respectively determining respective dissociation representation results of the shopping behavior sample set and the social behavior sample set through a preset model;
according to the dissociation characterization result of the shopping behavior sample set, correcting the dissociation characterization result of the social behavior sample set, and updating the model parameters of the preset model according to the correction result;
according to the dissociation characterization result of the social behavior sample set, correcting the dissociation characterization result of the shopping behavior sample set, and updating the model parameters of the preset model according to the correction result;
and after the preset model is iteratively updated for preset times, the commodity recommendation model is obtained.
3. The method of claim 2, wherein after determining the respective dissociation characterization results of the shopping behavior sample set and the social behavior sample set through a preset model, the method further comprises:
determining real influence parameters among the plurality of historical users according to respective dissociation representation results of the shopping behavior sample set and the social behavior sample set;
correcting the dissociation characterization result of the social behavior sample set according to the dissociation characterization result of the shopping behavior sample set, wherein the correction comprises the following steps:
correcting the dissociation characterization result of the social behavior sample set according to the real influence parameter and the social correction coefficient;
correcting the dissociation characterization result of the shopping behavior sample set according to the dissociation characterization result of the social behavior sample set, wherein the correction comprises the following steps:
and correcting the dissociation characterization result of the shopping behavior sample set according to the real influence parameters and the shopping correction coefficient.
4. The method of claim 3, wherein the dissociated representation of the sample set of shopping behavior comprises a plurality of first channels, each of the first channels representing a shopping intent of the user; the dissociation characterization result of the social behavior sample set comprises a plurality of second channels, and each second channel characterizes one social intention of the user;
determining real influence parameters among the plurality of historical users according to the respective dissociation characterization results of the shopping behavior sample set and the social behavior sample set, wherein the determination comprises the following steps:
respectively calculating the shopping representation similarity of the users of the plurality of first channels;
determining average shopping characterization similarity among a plurality of historical users according to the shopping characterization similarity of each user of the plurality of first channels;
respectively calculating the user social representation similarity of each of the plurality of second channels;
determining average social representation similarity among a plurality of historical users according to the user social representation similarity of each of the plurality of second channels;
and determining the real influence parameters according to the average shopping characterization similarity and the average social characterization similarity.
5. The method of claim 4, wherein the similarity of user shopping characteristics for each of the plurality of first channels is calculated by:
the calculation formula of the average shopping characterization similarity is as follows:
in the two formulas, X is the historical shopping behavior of the user;
is the k-thXThe shopping representation similarity of the users of the first channel; simXCharacterizing similarity for average purchases;
characterization for the k-th after dissociationXThe dissociation characterization results corresponding to the first channels; k is not less than 1X≤KX，KXIs the total number of channels of the first channel.
6. The method of claim 5, wherein the social representation similarity of the users of each of the plurality of second channels is calculated by:
the calculation formula of the average social representation similarity is as follows:
in the two formulas, A is the historical social behavior of the user;
is the k-thAThe social representation similarity of the users of the second channel; simACharacterize the similarity for average sociality;
characterization for the Kth after dissociationAThe dissociation characterization results corresponding to the second channels; k is not less than 1A≤KA，KAThe total number of channels of the second channel.
7. The method of claim 6, wherein the real impact parameter is calculated by:
sim＝simX·simA
in the formula, simXCharacterizing similarity for average purchases; simACharacterize the similarity for average sociality;
according to the real influence parameters and the shopping correction coefficients, a formula for correcting the dissociation characterization results of the shopping behavior sample set is as follows:
in the formula, alphaXCorrecting the coefficient for the purchase;
according to the real influence parameters and the social correction coefficient, a formula for correcting the dissociation characterization results of the social behavior sample set is as follows:
in the formula, alphaAIs a social correction factor.
8. An article recommendation device, the device comprising:
the acquisition module is used for acquiring historical shopping behaviors and social behaviors of a current user;
the vector representation module is used for respectively representing the historical shopping behaviors and the social behaviors of the current user by vectors;
the commodity recommendation module is used for inputting the vector representation of the historical shopping behaviors and the vector representation of the social behaviors of the current user into a pre-trained commodity recommendation model, and determining target commodities recommended to the current user through the pre-trained commodity recommendation model; the pre-trained commodity recommendation model is obtained by respectively performing dissociation representation on a shopping behavior sample set and a social behavior sample set based on shopping behavior sample sets and social behavior sample sets of a plurality of historical users and training according to an influence relation between a dissociation representation result of the shopping behavior sample set and a dissociation representation result of the social behavior sample set.
9. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor implements the item recommendation method of any one of claims 1 to 7 when executing the computer program.
10. A computer-readable storage medium, wherein a computer program is stored on the computer-readable storage medium, and when executed by a processor, the computer program implements the item recommendation method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210310761.1A CN114707063A (en) | 2022-03-28 | 2022-03-28 | Commodity recommendation method and device, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210310761.1A CN114707063A (en) | 2022-03-28 | 2022-03-28 | Commodity recommendation method and device, electronic equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114707063A true CN114707063A (en) | 2022-07-05 |
Family
ID=82171152
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210310761.1A Pending CN114707063A (en) | 2022-03-28 | 2022-03-28 | Commodity recommendation method and device, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114707063A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116205679A (en) * | 2023-02-27 | 2023-06-02 | 深圳市秦丝科技有限公司 | Physical store marketing recommendation method and device, electronic equipment and storage medium |
-
2022
- 2022-03-28 CN CN202210310761.1A patent/CN114707063A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116205679A (en) * | 2023-02-27 | 2023-06-02 | 深圳市秦丝科技有限公司 | Physical store marketing recommendation method and device, electronic equipment and storage medium |
| CN116205679B (en) * | 2023-02-27 | 2023-10-31 | 深圳市秦丝科技有限公司 | Physical store marketing recommendation method and device, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111460130B (en) | Information recommendation method, device, equipment and readable storage medium | |
| Merialdo | Clustering for collaborative filtering applications | |
| CN111797321B (en) | Personalized knowledge recommendation method and system for different scenes | |
| CN111310063B (en) | Neural network-based article recommendation method for memory perception gated factorization machine | |
| CN111061962B (en) | Recommendation method based on user scoring analysis | |
| CN109903103B (en) | Method and device for recommending articles | |
| CN108804577B (en) | Method for estimating interest degree of information tag | |
| CN112800207B (en) | Commodity information recommendation method and device and storage medium | |
| CN111768239A (en) | Property recommendation method, device, system, server and storage medium | |
| CN113761359B (en) | Data packet recommendation method, device, electronic equipment and storage medium | |
| CN113592609A (en) | Personalized clothing matching recommendation method and system using time factors | |
| CN111782927A (en) | Article recommendation method and device, computer storage medium | |
| CN112380449A (en) | Information recommendation method, model training method and related device | |
| CN114707063A (en) | Commodity recommendation method and device, electronic equipment and storage medium | |
| Zhang et al. | Cross-domain recommendation with multiple sources | |
| CN113641811B (en) | Session recommendation method, system, equipment and storage medium for promoting purchasing behavior | |
| CN113592593A (en) | Training and application method, device, equipment and storage medium of sequence recommendation model | |
| CN113688306A (en) | Recommendation strategy generation method and device based on reinforcement learning | |
| CN115730217A (en) | Model training method, material recalling method and device | |
| CN115809374A (en) | Method, system, device and storage medium for correcting mainstream deviation of recommendation system | |
| CN113254788B (en) | Big data based recommendation method and system and readable storage medium | |
| CN112734519B (en) | Commodity recommendation method based on convolution self-encoder network | |
| CN115510318A (en) | Training method of user characterization model, user characterization method and device | |
| CN114357283A (en) | Recommendation method based on comment and self-attention mechanism | |
| CN113065067A (en) | Article recommendation method and device, computer equipment and storage medium |
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 |