CN117934081A

CN117934081A - Content recommendation method, system, electronic device and storage medium

Info

Publication number: CN117934081A
Application number: CN202311863421.2A
Authority: CN
Inventors: 刘晓宏; 肖智国; 王攀登; 蒋恒伟; 江爱琼; 张龙
Original assignee: Baweitong Technology Co ltd
Current assignee: Baweitong Technology Co ltd
Priority date: 2023-12-29
Filing date: 2023-12-29
Publication date: 2024-04-26

Abstract

The application provides a content recommendation method, a content recommendation system, electronic equipment and a storage medium, and belongs to the technical field of intelligent recommendation. The method includes acquiring a user profile of a target user when an opportunity point event is detected to be triggered. Meanwhile, the content information to be recommended is selected from the content information base, a plurality of recommendation algorithm combinations are adopted, and the matching degree of each recommended content in the content information to be recommended and the user portrait is calculated, so that the recommended content with the highest matching degree with the user portrait is the target recommended content, and is pushed to the target user. Because the content information to be recommended comprises the content information related to the trip information of the target user and the content information related to the opportunity point event, the content can be effectively recommended aiming at the user portrait and the user trip information, and the recommendation accuracy can be improved.

Description

Content recommendation method, system, electronic device and storage medium

Technical Field

The present application relates to the field of intelligent recommendation technologies, and in particular, to a content recommendation method, system, electronic device, and storage medium.

Background

The riding APP, such as the soe line, is convenient for users to inquire related information by taking subways, and can recommend diversified information contents for users to select for use, so that various demands of the users are met. In the process of providing services for users, how to effectively recommend content to users and travel information of the users becomes a problem to be solved.

Disclosure of Invention

The embodiment of the application mainly aims to provide a content recommendation method, a system, electronic equipment and a storage medium, aiming at obtaining the matching degree of recommended content containing content information related to travel information and content information related to the opportunity point event and a user portrait by adopting a plurality of recommendation algorithm combination calculation when the opportunity point event is detected to be triggered, so that content recommendation can be effectively carried out on the user portrait and the user travel information, and the recommendation accuracy can be improved.

To achieve the above object, a first aspect of an embodiment of the present application provides a content recommendation method, including:

when the opportunity point event is detected to be triggered, acquiring a user portrait of a target user, wherein the opportunity point event is an event which has an opportunity to recommend to the target user;

Selecting content information to be recommended from a content information base, wherein the content information to be recommended comprises content information related to trip information of the target user and content information related to the opportunity point event;

Calculating the matching degree of each recommended content in the content information to be recommended and the user portrait by adopting a plurality of recommendation algorithm combinations;

And taking the recommended content with the highest matching degree with the user portrait as target recommended content, and pushing the target recommended content to the target user.

In one embodiment of the application, after pushing the target recommended content to the target user, the method comprises:

acquiring feedback information of the target user on the target recommended content;

determining whether the recommendation of the target recommended content is accurate or not according to the feedback information;

and when the inaccuracy of the recommendation of the target recommended content is determined, optimizing the user portrait according to the feedback information and optimizing the weights of a plurality of recommendation algorithms, returning to the step of calculating the matching degree of each recommended content in the content information to be recommended and the user portrait by adopting a plurality of recommendation algorithm combinations.

In one embodiment of the present application, the selecting content information to be recommended from the content information base includes:

According to the user portrait, first content information of which the target user is an audience group is screened out from the content information base, wherein the first content information comprises product information, activity information, advertisement information and service information;

screening second content information related to the travel information from the content information base according to the travel information of the target user;

Screening third content information related to the opportunity point event from the content information base according to the opportunity point event;

And taking the first content information, the second content information and the third content information as content information to be recommended.

In one embodiment of the application, constructing the user representation comprises:

Determining a first user portrait according to the attribute data and the historical behavior data of the target user;

Acquiring travel information of the target user, analyzing the travel information, and determining potential requirements of the target user in the travel process;

And optimizing the first user portrait according to the potential requirement of the target user in the traveling process, and constructing the user portrait.

In one embodiment of the application, the method further comprises:

acquiring user portrait and travel information of the target user;

And selecting target recommended content which is matched with the user portrait and related to the travel information from the content information base by utilizing a plurality of recommendation algorithm combinations, and actively pushing the target recommended content to the target user.

In one embodiment of the application, the method further comprises:

determining target recommended content to be recommended from a content information base;

obtaining a target user portrait matched with the target recommended content by utilizing a plurality of recommendation algorithm combinations;

and pushing the target recommended content to a target user group corresponding to the target user image, wherein the target user group comprises at least one target user.

In one embodiment of the present application, the calculating, by using a combination of a plurality of recommendation algorithms, a matching degree between each recommended content in the content information to be recommended and the user portrait includes:

determining weights corresponding to various recommendation algorithms according to the perfection of the user portrait;

And calculating the matching degree of each recommended content in the content information to be recommended and the user portrait by using various recommendation algorithms and the corresponding weights thereof.

A second aspect of an embodiment of the present application proposes a content recommendation system, including: the system comprises a user portrait library, a content information library and a recommendation module, wherein the user portrait library and the content information library are connected with the recommendation module;

The user portrait library is used for managing and optimizing user portraits of all users;

The content information base is used for managing each content information;

the recommending module is used for executing the recommending method according to any embodiment of the application.

A third aspect of the embodiments of the present application proposes an electronic device comprising a memory, a processor, a program stored on the memory and executable on the processor, and a data bus for enabling a connected communication between the processor and the memory, the program when executed by the processor implementing the steps of the recommended method according to any of the embodiments of the present application.

A fourth aspect of the embodiments of the present application proposes a storage medium, which is a computer-readable storage medium, for computer-readable storage, where one or more programs are stored, and the one or more programs are executable by one or more processors to implement the steps of the recommendation method according to any of the embodiments of the present application.

The application provides a content recommendation method, a content recommendation system, electronic equipment and a storage medium. Meanwhile, the content information to be recommended is selected from the content information base, a plurality of recommendation algorithm combinations are adopted, and the matching degree of each recommended content in the content information to be recommended and the user portrait is calculated, so that the recommended content with the highest matching degree with the user portrait is the target recommended content, and is pushed to the target user. Because the content information to be recommended comprises the content information related to the trip information of the target user and the content information related to the opportunity point event, the content can be effectively recommended aiming at the user portrait and the user trip information, and the recommendation accuracy can be improved.

Drawings

FIG. 1 is a flow chart of a content recommendation method provided in an embodiment of the present application;

FIG. 2 is a flowchart of the steps for constructing a user representation provided by an embodiment of the present application;

FIG. 3 is a flowchart illustrating steps for selecting content information to be recommended from a content information base according to an embodiment of the present application;

FIG. 4 is a flowchart of a step of calculating the matching degree between each recommended content in the content information to be recommended and the user portrait by adopting a plurality of recommendation algorithm combinations according to the embodiment of the application;

FIG. 5 is a flowchart of steps performed after pushing target recommended content to a target user provided by an embodiment of the present application;

FIG. 6 is another flowchart of a content recommendation method provided by an embodiment of the present application;

FIG. 7 is another flowchart of a content recommendation method provided by an embodiment of the present application;

FIG. 8 is another flowchart of a content recommendation method provided by an embodiment of the present application;

FIG. 9 is another flowchart of a content recommendation method provided by an embodiment of the present application;

FIG. 10 is another flowchart of a content recommendation method provided by an embodiment of the present application;

FIG. 11 is a diagram illustrating a recommendation flow in a business scenario provided by an embodiment of the present application;

FIG. 12 is a schematic flow chart of active recommendation provided by an embodiment of the present application;

FIG. 13 is a schematic flow chart of passive recommendation provided by an embodiment of the present application;

FIG. 14 is a detailed flow diagram of passive recommendation provided by an embodiment of the present application;

FIG. 15 is a schematic diagram of a content recommendation system according to an embodiment of the present application;

fig. 16 is a schematic diagram of a hardware structure of an electronic device according to an embodiment of the present application.

Detailed Description

The present application will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present application more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the application.

It should be noted that although functional block division is performed in a device diagram and a logic sequence is shown in a flowchart, in some cases, the steps shown or described may be performed in a different order than the block division in the device, or in the flowchart. The terms first, second and the like in the description and in the claims and in the above-described figures, are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order.

Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of the application only and is not intended to be limiting of the application.

With the development of internet technology, people browse content information provided by a network platform more and more frequently. For example, merchandise information is browsed in an online shopping platform, hot spot information is browsed in a news platform, financial information is browsed in a financial platform, payment service information is browsed in a payment platform, and the like.

Different users have more or less different demands for their provided content information when using the same network platform. On the other hand, the massive growth of information in network platforms also often makes it difficult for users to choose. At present, content information recommended to a user is insufficient in accuracy and the like, so that personalized requirements of the user are difficult to meet.

For example, riding APP is convenient for users to inquire related information by subway, and meanwhile, diversified information contents can be recommended for users to select for use, so that various demands of the users are met. In the process of providing services for users, how to effectively recommend content to users and travel information of the users becomes a problem to be solved.

Based on the above, the embodiment of the application provides a content recommendation method, a system, an electronic device and a storage medium, which aim to obtain the matching degree of the recommended content containing the content information related to the trip information and the content information related to the trip event and the user portrait by adopting a plurality of recommendation algorithm combinations when the trip event is triggered, so that the content recommendation can be effectively performed on the user portrait and the trip information of the user, and the recommendation accuracy can be improved.

Referring to fig. 1, fig. 1 is a flowchart of a content recommendation method provided in an embodiment of the present application, and the method in fig. 1 may include, but is not limited to, steps S110 to S140.

Step S110, when the event of the opportunity point is detected to be triggered, the user portrait of the target user is acquired, and the event of the opportunity point is an event which has an opportunity to recommend to the target user.

In the embodiment of the application, the opportunity point event is an event which has the opportunity to recommend to the target user. The opportunity point events include inbound events, outbound events, approach business turn events, card line shortage events, point events, etc., which originate from business system message triggers. Specifically, the entry event is an entry event that is triggered when a passenger enters a subway station by using a two-dimensional code of a riding App or a face or the like. The inbound event typically also includes the following information: passenger attribute data, account numbers, ticket numbers, time and location information of inbound stops, inbound stop names, etc. The outbound event is triggered by using a two-dimensional code of a riding App or a face when a passenger goes out of a subway station. The outbound event typically also includes the following information: passenger attribute data, account numbers, ticket numbers, outbound time and outbound site location information, outbound site names, etc. The route business district event is triggered by the route business district or shopping area according to the geographical position information sent by the user in the journey of the passenger taking the subway. The event of insufficient credit card amount is when passengers pay vehicles by using the balance of subway traffic cards or two-class and three-class bank accounts, and if the balance in the card is insufficient to pay the whole-course fee, the event of insufficient credit card amount is triggered. The integration event is generated when a passenger takes a car, and the integration event is triggered when the integration reaches a certain amount. In the embodiment of the application, when the event of the opportunity point is detected to be triggered, the user portrait of the target user is acquired, so that corresponding recommendation can be performed according to the matching degree of the user portrait and the information to be recommended.

In one embodiment of the present application, referring to fig. 2, fig. 2 is a flowchart of steps for constructing a user portrait according to an embodiment of the present application, including but not limited to steps S210 to S230.

Step S210, determining a first user portrait according to attribute data and historical behavior data of a target user;

Step S220, obtaining travel information of a target user, analyzing the travel information and determining potential requirements of the target user in the travel process;

and step S230, optimizing the first user portrait according to the potential requirements of the target user in the traveling process, and constructing the first user portrait.

In the embodiment of the present application, the user attribute data is also called user demographic data, that is, the attribute of the user, such as age, gender, region, academic, family composition, occupation, etc. These data are generally steady (e.g., gender) or slowly (e.g., age). The human being is a socialization species, different attributes of the user determine that the user is in different levels or life circles, the different levels or life circles have different behavior characteristics, life patterns and preference characteristics, and the same circle layer has certain similarity, so that the similarity provides certain guarantee for personalized recommendation. The historical behavior data of the user is all operations of the user on the product, such as browsing, clicking, playing, purchasing, searching, collecting, praying, forwarding, adding shopping carts, even sliding, staying at a certain position, fast forwarding and the like. The user's operational behavior is feedback of the user's most realistic intent, which reflects the user's interest state, and by analyzing the user's behavior, deep insight into the user's interest preferences can be obtained. User behavior is generally classified into explicit behavior and implicit behavior depending on whether the user's behavior directly indicates the user's preference for interest in the target item. Explicit behavior is behavior that directly indicates the user's interests, such as praise, scoring, etc. Implicit actions, while not directly representing the user's interests, may indirectly feed back the user's interest changes, so long as the user does not score directly, click, play, collect, comment, forward, etc., all operate to calculate implicit feedback. User portrayal, namely user information tagging, is to perfectly abstract the business full view of a user as a basic mode of applying big data technology after collecting and analyzing data of main information such as user social attribute, living habit, consumption behavior and the like. The user image can provide enough information foundation, and can help to quickly find out more extensive feedback information such as accurate user groups and user demands. The focus of the user representation is to "tag" the user, and one tag is usually a manually specified highly refined feature identifier, such as age, gender, region, user preference, etc., and finally the three-dimensional "representation" of the user can be outlined by combining all the tags of the user. In the embodiment of the application, the preliminary user portrait of the user, namely the first user portrait, can be constructed and obtained based on the attribute data and the historical behavior data of the target user. Meanwhile, based on the characteristics of the riding APP, the traveling information of the target user can be obtained through the riding APP, so that the traveling information of the target user can be analyzed according to the traveling information of the target user, and the potential requirements of the target user in the traveling process can be determined. Therefore, the first user portrait is optimized according to the potential demands of the target user in the traveling process, and the user portrait is constructed and obtained, so that the user portrait is more comprehensive.

Travel information such as departure places and destinations of users is properly used, the data are integrated into a recommendation algorithm, and personalized recommendation can be performed for the users more accurately and in a scene.

In the embodiments of the present application, when related processing is performed according to user information, user behavior data, user history data, user location information, and other data related to user identity or characteristics, permission or consent of the user is obtained first, and the collection, use, processing, and the like of the data comply with related laws and regulations and standards of related countries and regions. In addition, when the embodiment of the application needs to acquire the sensitive personal information of the user, the independent permission or independent consent of the user is acquired through popup or jump to a confirmation page and the like, and after the independent permission or independent consent of the user is definitely acquired, the necessary relevant data of the user for enabling the embodiment of the application to normally operate is acquired.

Step S120, selecting content information to be recommended from the content information base, wherein the content information to be recommended comprises content information related to travel information of the target user and content information related to the opportunity point event.

In its own embodiment, the content information base includes all content information that can be recommended, including product information, campaign information, advertisement information, service information, and the like. Wherein, each product information corresponds to a product label, such as a product category, a product function characteristic, a product brand, a product applicable crowd, and the like. The activity information also corresponds to activity labels, such as activity types, activity topics, activity time periods, activity participation objects and the like. The advertisement information also corresponds to advertisement labels, such as advertisement content, advertisement form, advertisement audience, etc. The service information includes service items, service audiences, and the like. Such as parking service information, which corresponds to parking service tags, such as parking locations, prices, facilities, security, etc. And preliminarily selecting content information which is suitable for being recommended to the target user from the content information base as content information to be recommended. The information to be recommended needs to contain content information related to trip information of the target user and content information related to the opportunity point event. For example, the destination of the target user can be obtained through the trip information, so that the information to be recommended also needs to include advertisement information, activity information, service information and the like related to the destination, such as clothing advertisements in the market at the destination, diet restaurant information with discount offers, parking lot information with offers and the like. Meanwhile, the information to be recommended needs to include content information related to the opportunity point event, for example, when the event of insufficient credit is triggered, the recommended content information needs to include content recommendations of other payment modes. When the approach business district event is triggered, the recommended content information needs to include preferential activity content recommendation, preferential restaurant information recommendation and the like of the approach business district.

In one embodiment of the present application, referring to fig. 3, fig. 3 is a flowchart of a step of selecting content information to be recommended from a content information base according to an embodiment of the present application, including but not limited to steps S310 to S340.

Step S310, first content information of which the target user is an audience group is screened out from a content information base according to the user portrait, wherein the first content information comprises product information, activity information, advertisement information and service information;

Step S320, second content information related to the travel information is screened out from the content information base according to the travel information of the target user;

Step S330, screening out third content information related to the opportunity point event from the content information base according to the opportunity point event;

in step S340, the first content information, the second content information, and the third content information are used as the content information to be recommended.

In the embodiment of the application, the content information base comprises all the content information which can be recommended, including product information, activity information, advertisement information, service information and the like. Therefore, first content information of the target user as an audience group needs to be screened from the content information base according to the user portrait. Specifically, each piece of content information has a corresponding content tag. The user portrait also contains the user tag. By matching the content tag with the user tag, first content information of which the target user is an audience group can be screened out from the content information base. Wherein the first content information includes product information, campaign information, advertisement information, and service information. Meanwhile, second content information related to the travel information is screened out from the content information base according to the travel information of the target user. And screening third content information related to the opportunity point event from the content information base according to the opportunity point event. The first content information, the second content information and the third content information which are screened out can be used as the content information to be recommended.

In the embodiment of the application, the information to be recommended, which is screened out from the content information base, not only comprises the first content information of which the target user is an audience group and the third content information related to the opportunity point event, but also comprises the second content information related to the trip information, so that the requirements of the target user can be timely acquired in the trip process of the user, the targeted content recommendation can be carried out, and the recommendation accuracy can be improved.

And step S130, combining a plurality of recommendation algorithms, and calculating to obtain the matching degree of each recommended content in the content information to be recommended and the user portrait.

In the embodiment of the application, the plurality of recommendation algorithms comprise an accurate matching algorithm, a collaborative filtering algorithm, a popular recommendation algorithm, a matrix decomposition algorithm and the like, wherein the accurate matching algorithm is used for carrying out content label matching according to labels of users. Collaborative filtering algorithms are based on the recommendation of a user to whom a user likes a similar subject matter (which the user has not operated), i.e., collaborative filtering based on items (subject matter), which is easy to process on a large scale data set by adding computing nodes. It has proven to be very good in practical use to recommend diverse, novel targets to users. Especially when the population size is larger, the more the user acts, the better the recommended effect. The popular recommendation is to recommend the content information with the largest popularity, such as transaction amount, browsing amount, click amount, sharing amount, etc., to the current content information. According to the embodiment of the application, a plurality of recommendation combination algorithms are adopted, weight distribution of the recommendation algorithms is carried out according to different conditions, and then the matching degree of each recommendation content in the content information to be recommended and the user portrait is calculated.

In the embodiment of the application, a collaborative filtering algorithm based on articles can be adopted for recommendation, and the specific process is as follows:

1. Constructing a user-item scoring matrix: the scoring information of the user is represented as a matrix, wherein the rows represent the user and the columns represent the items, and the elements of the matrix are the user's scores for the items.

2. Calculating the similarity of articles: similarity between items is calculated using some similarity measure. This may be accomplished by calculating the association rules of the items or using vector similarity (e.g., cosine similarity), etc.

3. Finding similar items: and selecting some articles with highest similarity to the target articles to construct a similar article set.

4. Generating a recommendation: non-scored items that may be of interest are recommended to the target user based on the user's historical preferences for similar items.

5. Filtering the scored items: items that the target user has scored are filtered from the recommendation list.

For example, assuming there are N items, each item may be represented as a vector, the dimensions of the vector corresponding to the characteristics of the item. The similarity between the items is calculated, and in particular, a different similarity measure, such as cosine similarity or pearson correlation coefficient, may be used for each pair of items i and j. Let the similarity matrix be S, where S _ij represents the similarity between item i and item j. Generating a recommendation list for the user U, in particular, for items i that the user U does not interact with, the recommendation list may be generated by:

a. for item j that user U has interacted with, find the most similar set of non-interacted items to j.

B. based on the similarity and the user's score for item j, a score is predicted for item i that user U did not interact with, e.g., by weighting the similarity:

Where N (U) represents the set of items interacted by user U, S (U, i) represents the predicted score of user U for item i that did not interacted, S (U, j) represents the score of user U for item i that interacted, and S _ij represents the similarity between item i and item j.

In this way, scores of all items which are not interacted with the user U can be predicted, and then the predicted scores are ranked to generate a corresponding item recommendation list.

In the embodiment of the application, because the interests of the users are relatively continuous and the characteristics of the articles are relatively static, the similarity change among the articles is relatively small, and therefore, compared with the collaborative filtering algorithm based on the users, the collaborative filtering algorithm based on the articles is more stable. Moreover, with a large number of items, collaborative filtering based on items is generally computationally less complex and can be better extended to large-scale systems. For items that are newly added to the system, the item-based collaborative filtering algorithm is also more advantageous because the user's historical behavior has less impact on the recommendation of items. And in the case of a large number of items but a small number of users, it is easier to calculate the similarity between items than to calculate the similarity between users.

Of course, collaborative filtering algorithms based on items also have some drawbacks, such as in the case of a large number of users but a small number of items, calculating the similarity between items may become complicated and the calculation amount is relatively large. For new users, it is difficult to provide accurate recommendations for them, i.e. cold start problems, due to lack of user historical behavioral data. At the same time there may be a problem with item popularity preferences, i.e. recommending items that are already very popular, while ignoring some potentially interesting but relatively cool items. And the user-item scoring matrix is typically sparse, i.e., the user scores only a few items. This results in a difficulty in accurately calculating the similarity between the articles, especially in the case of sparse data. Based on the above, in the actual service scene, the embodiment of the application optimizes the problem of the algorithm in the service, gathers data on a large scale of articles, reduces the complexity of calculation by using a dimension reduction technology (such as singular value decomposition), introduces time factors, and pays more attention to recent user behaviors by using a weighting mode so as to adapt to the variation of the popularity of the articles. When the user behavior is changed, the model is updated in an incremental updating mode, and the whole model is prevented from being recalculated each time.

In the embodiment of the application, a matrix decomposition algorithm can be adopted for recommendation, and the specific process is as follows:

1. initializing: the user matrix U and the item matrix V are randomly initialized. Typically, the dimensions of the matrix are determined by the hyper-parameters.

2. Loss function definition: an optimized loss function, typically Root Mean Square Error (RMSE) between the score predictor and the actual score, is defined, possibly with the addition of regularization terms.

3. Gradient descent or optimization algorithm: optimization algorithms such as gradient descent are employed to minimize the loss function. For each user-item pair, a gradient is calculated and the user matrix U and the item matrix V are updated.

4. Iterative optimization: the gradient descent step is repeatedly performed until the loss function converges or a predetermined number of iterations is reached.

5. And (3) loss function convergence judgment: and monitoring the convergence condition of the loss function, and stopping training if the convergence condition is met.

6. Model output: an optimal user matrix U and an item matrix V are obtained, and the product of the two matrices approximately restores the original scoring matrix.

7. Recommendation generation: using the learned user matrix U and item matrix V, the user's scoring of the unscored items may be predicted, thereby generating a personalized recommendation list.

8. Model evaluation: and evaluating the model by using a verification set or a test set, and calculating the error between the predicted score and the true score.

9. Super parameter tuning: depending on the evaluation of the model, it may be necessary to adjust super parameters, such as learning rate, regularization parameters, etc., to improve the model performance.

10. Model application: and applying the trained model to an actual scene, and providing personalized recommendation for the user.

Illustratively, a user-item scoring matrix R is given and singular value decomposition is performed to decompose it into the product of three matrices: r=uΣv ^T;

Wherein U is an orthogonal matrix, and its column vector is the characteristic vector of the user; Σ is a diagonal matrix whose diagonal elements are singular values; v is an orthogonal matrix whose row vector is the feature vector of the article; t represents the transpose of the matrix.

Calculating a user-item scoring matrix R: r=uΣv ^T;

solving U: u=r·v·Σ ^-1;

solving V: v=r ^T·U·Σ^-1;

Solving for Σ:

Where Σ ^-1 represents the inverse of the matrix Σ.

In the embodiment of the application, the potential characteristics of the user and the object can be learned through decomposing the matrix. These features may capture the relationship between the user and the item, making the recommendation more personalized. And the user-item scoring matrix in the recommendation system is usually sparse, i.e. most items are missing, and matrix decomposition has better adaptability for processing the sparsity. Meanwhile, implicit features learned through the decomposition matrix have a certain interpretability, so that the user and the article can be helped to understand the association. Matrix decomposition algorithms can be extended to large-scale systems, typically by means of distributed computing, etc., and are suitable for handling large numbers of users and items.

Of course, matrix factorization algorithms also have some drawbacks, such as for new users or items, the matrix factorization may not accurately predict their interests due to the lack of historical scoring information. Models are prone to overfitting when noise or outliers are present in the scoring matrix. To mitigate over-fitting, it is often necessary to introduce regularization or the like. In large-scale systems, computing the decomposition of the user-item scoring matrix may become more complex, especially in the context of real-time recommendations. Matrix factorization is a linear model that makes it difficult to capture the nonlinear relationship between a user and an item. Some of the super parameters in the matrix decomposition algorithm usually need to be optimized, which may cause some trouble for non-professionals. Based on the above, in the actual service scene, the embodiment of the application introduces regularization term aiming at the over-fitting problem of the algorithm in the service, prevents the model from fitting training data excessively and improves generalization capability. Implicit feedback information of the user, such as clicking, browsing records and the like of the user, is introduced, the interest of the user is known in all directions, and therefore model parameters are dynamically adjusted, and a deep learning model (deep matrix decomposition) is used for capturing nonlinear relations between the user and the object better.

Referring to fig. 4, fig. 4 is a flowchart illustrating steps of calculating a matching degree between each recommended content in the content information to be recommended and the user portrait by using a combination of multiple recommendation algorithms according to an embodiment of the present application, including but not limited to steps S410 to S420.

Step S410, determining the weights corresponding to various recommendation algorithms according to the perfection of the user portrait;

step S420, calculating and obtaining the matching degree of each recommended content in the content information to be recommended and the user portrait by using various recommendation algorithms and the corresponding weights.

In the embodiment of the application, the weights corresponding to various recommendation algorithms can be determined according to the perfection of the user portrait. For example, when the historical behavior data of the user is less, the user portrait is not perfect enough, namely the user tag is not perfect enough, and at the moment, the recommendation effect of the accurate matching algorithm is not good. In this case, the weight ratio of the exact matching algorithm needs to be properly reduced, and the weight ratio of the popular recommendation algorithm needs to be improved to achieve a better recommendation effect. When a user generates a large amount of behavior data, in order to improve the richness and the accuracy of recommendation, a user similarity matrix and a product similarity matrix can be constructed through historical data to recommend commercial content, so that the purpose of accurately meeting user preferences is achieved. I.e. the weight ratio of the collaborative filtering algorithm needs to be properly increased. And a large amount of data is generated along with the user behavior, and a more accurate user label is provided, so that the recommendation is performed by using an accurate matching algorithm, and at the moment, the better recommendation effect can be achieved by improving the weight proportion of the accurate matching algorithm. In the embodiment of the application, the weights corresponding to various recommendation algorithms are determined according to the perfection of the user portrait. Therefore, the matching degree of each recommended content in the content information to be recommended and the user portrait can be calculated by utilizing various recommendation algorithms and the corresponding weights. According to the embodiment of the application, through effective matching of a plurality of recommendation algorithms, the problems of a single algorithm can be avoided, and the overall quality of recommendation is improved. The user can obtain high-quality recommendation service in real time, and the user can be helped to obtain useful information in time.

And step S140, taking the recommended content with the highest matching degree with the user image as the target recommended content, and pushing the target recommended content to the target user.

In the embodiment of the application, after the matching degree of each recommended content in the content information to be recommended and the user portrait is calculated by utilizing a plurality of recommendation algorithm combinations, the recommended content with the highest matching degree with the user portrait is taken as the target recommended content, and the target recommended content is pushed to the target user.

Illustratively, the content information to be recommended includes 5 recommended contents including recommended content a, recommended content B, recommended content C, recommended content D, and recommended content E. And calculating the matching degree between the recommended content A and the user portrait to be E1 by utilizing a plurality of recommendation algorithm combinations. And calculating to obtain the matching degree E2 of the recommended content B and the user portrait by utilizing a plurality of recommendation algorithm combinations. And calculating to obtain the matching degree E3 of the recommended content C and the user portrait by utilizing the combination of a plurality of recommendation algorithms. And calculating to obtain the matching degree E4 of the recommended content D and the user image by utilizing a plurality of recommendation algorithm combinations. And calculating to obtain the matching degree E5 of the recommended content E and the user portrait by utilizing the combination of a plurality of recommendation algorithms. The matching degree E5 is highest, so that the recommended content E can be pushed to the target user.

In the embodiment of the application, the content information to be recommended comprises the content information related to the trip information of the target user and the content information related to the opportunity point event, so that effective content recommendation can be performed on the user portrait and the trip information of the user, and the recommendation accuracy can be improved.

In one embodiment of the present application, referring to fig. 5, fig. 5 is a flowchart of steps performed after pushing target recommended content to a target user, including but not limited to steps S510 to S530, provided by an embodiment of the present application.

Step S510, obtaining feedback information of a target user on target recommended content;

step S520, determining whether the recommendation of the target recommended content is accurate or not according to the feedback information;

and step S530, when the inaccuracy of the recommendation of the target recommended content is determined, optimizing the user portrait according to the feedback information and optimizing the weights of a plurality of recommendation algorithms, and returning to the step of calculating the matching degree of each recommended content in the content information to be recommended and the user portrait by adopting the combination of the plurality of recommendation algorithms.

In the embodiment of the application, after the target recommended content is recommended to the target user by utilizing the combination of a plurality of recommendation algorithms, the feedback information of the target user on the target recommended content can be further acquired, so that whether the recommendation of the target recommended content is accurate or not can be determined according to the feedback information. The feedback information comprises click, comment, sharing, forwarding, success rate and other behavior data of the user. And when the inaccuracy of the recommendation of the target recommended content is determined, optimizing the user portrait according to the feedback information and optimizing the weights of a plurality of recommendation algorithms, and returning to the step of calculating the matching degree of each recommended content in the content information to be recommended and the user portrait by adopting the combination of the plurality of recommendation algorithms. When the recommendation of the target recommended content is determined to be accurate, the recommendation is continuously performed in the follow-up recommendation.

Referring to fig. 6, fig. 6 is another flowchart of a content recommendation method according to an embodiment of the present application. Including but not limited to steps S610 through S680.

Step S610, when the event of the opportunity point is detected to be triggered, the user portrait of the target user is obtained, and the event of the opportunity point is an event which has an opportunity to recommend to the target user;

step S620, selecting content information to be recommended from a content information base, wherein the content information to be recommended comprises content information related to travel information of a target user and content information related to an opportunity point event;

Step S630, adopting a plurality of recommendation algorithm combinations to calculate and obtain the matching degree of each recommended content in the content information to be recommended and the user portrait;

step S640, taking the recommended content with the highest matching degree with the user image as the target recommended content, and pushing the target recommended content to the target user;

Step S650, obtaining feedback information of a target user on target recommended content;

step S660, judging whether the recommendation is accurate or not according to the feedback information;

step S670, when the recommendation is accurate, continuing the recommendation;

in step S680, when the recommendation is inaccurate, the user portrait is optimized according to the feedback information and the weights of the various recommendation algorithms are optimized, and then the process returns to step S630.

In the embodiment of the application, for each passive recommendation triggered by the opportunity point event, whether the recommendation is accurate or not can be judged by acquiring the feedback information of the target user on the target recommendation content. Therefore, when the recommendation is inaccurate, the user portrait can be optimized according to the feedback information, and the weights of various recommendation algorithms can be optimized. And then, based on the optimized user portrait and the weights of various recommendation algorithms, the matching degree of each recommended content in the content information to be recommended and the user portrait can be recalculated so as to be recommended again. The recommendation accuracy can be improved by continuously optimizing the user portraits and continuously optimizing the weights of various recommendation algorithms.

In one embodiment of the present application, referring to fig. 7, fig. 7 is another flowchart of a content recommendation method provided in an embodiment of the present application, including but not limited to steps S710 to S720.

Step S710, obtaining user portrait and travel information of a target user;

And step S720, selecting target recommended content which is matched with the user portrait and related to trip information from the content information base by utilizing a plurality of recommendation algorithm combinations, and actively pushing the target recommended content to a target user.

In the embodiment of the application, the active recommendation can be performed at any time and any place besides the passive recommendation performed when the opportunity point event is triggered. Specifically, the user portrait and trip information of the target user are obtained, so that a plurality of recommendation algorithm combinations can be utilized to select target recommendation contents which are matched with the user portrait and related to the trip information from a content information base to be actively pushed to the target user.

Illustratively, some needs of the target user may be determined based on the user profile of the target user. For example, according to the user portrait of the user, the content information a, the content information B and the content information C which may be of interest to the target user are determined. The travel information of the target user is acquired, information such as a departure place, a route business district, a destination and the like of the target user can be determined, and the content information related to the travel information is selected from the content information A, the content information B and the content information C to be the content information A, so that the content information A can be recommended to the target user.

Referring to fig. 8, fig. 8 is another flowchart of a content recommendation method according to an embodiment of the present application, including but not limited to steps S810 to S860.

Step S810, obtaining user portrait and travel information of a target user;

Step S820, selecting target recommended content which is matched with the user portrait and related to travel information from a content information base by utilizing a plurality of recommendation algorithm combinations, and actively pushing the target recommended content to a target user;

step S830, obtaining feedback information of a target user on target recommended content;

step S840, judging whether the recommendation is accurate or not according to the feedback information;

step S850, continuously recommending when the recommendation is accurate;

step S860, when the recommendation is inaccurate, the user portrait is optimized according to the feedback information and the weights of the plurality of recommendation algorithms are optimized, and then the step S810 is returned.

In the embodiment of the application, for active pushing, whether the recommendation is accurate can be judged by acquiring the feedback information of the target user on the target recommended content. Therefore, when the recommendation is inaccurate, the user portrait can be optimized according to the feedback information, and the weights of various recommendation algorithms can be optimized. And then, based on the optimized user portrait and the weights of various recommendation algorithms, the matching degree of each recommended content in the content information to be recommended and the user portrait can be recalculated so as to be recommended again. The recommendation accuracy can be improved by continuously optimizing the user portraits and continuously optimizing the weights of various recommendation algorithms.

In one embodiment of the present application, referring to fig. 9, fig. 9 is another flowchart of a content recommendation method provided in an embodiment of the present application, including but not limited to steps S910 to S930.

Step S910, determining target recommended content to be recommended from a content information base;

Step S920, a target user portrait matched with target recommended content is obtained by utilizing a plurality of recommendation algorithm combinations;

In step S930, the target recommended content is pushed to the target user group corresponding to the target user image, where the target user group includes at least one target user.

In the embodiment of the application, the active recommendation can be performed at any time and any place besides the passive recommendation performed when the opportunity point event is triggered. Specifically, in addition to the fact that the user portrait and trip information of the target user can be determined first to match related content information, target recommended content to be pushed can be determined first, and then the target recommended content matches a corresponding target user group. Thus, all target users in the target user group can be recommended to the target recommended content. Specifically, multiple recommendation algorithm combinations are needed to obtain the target user portraits matched with the target recommended content, so that the target recommended content can be pushed to the target user group corresponding to the target user portraits. Wherein the target user group comprises at least one target user.

The user image library is managed to determine the target user group corresponding to the target user image. The target user group includes one or more target users. Thus, after the target user portraits corresponding to the target recommended content are determined by utilizing the combination of a plurality of recommendation algorithms, the target recommended content can be recommended to each target user in the target user group based on the target user group corresponding to the target user portraits.

Referring to fig. 10, fig. 10 is another flowchart of a content recommendation method according to an embodiment of the present application, including but not limited to steps S1010 to S1070.

Step S1010, determining target recommended content to be recommended from a content information base;

Step S1020, a target user portrait matched with target recommended content is obtained by utilizing a plurality of recommendation algorithm combinations;

Step S1030, pushing target recommended content to a target user group corresponding to the target user image, wherein the target user group comprises at least one target user;

Step S1040, obtaining feedback information of a target user group on target recommended content;

step S1050, judging whether the recommendation is accurate or not according to the feedback information;

Step S1060, when the recommendation is accurate, the recommendation is continued;

Step S1070, when the recommendation is inaccurate, the user portrait is optimized according to the feedback information and the weights of a plurality of recommendation algorithms are optimized, and then the step S1020 is returned.

In the embodiment of the application, for active pushing, whether the recommendation is accurate can be judged by acquiring the feedback information of the target user group on the target recommended content. Therefore, when the recommendation is inaccurate, the user portrait can be optimized according to the feedback information, and the weights of various recommendation algorithms can be optimized. And then, based on the optimized user portrait and the weights of various recommendation algorithms, the matching degree of each recommended content in the content information to be recommended and the user portrait can be recalculated so as to be recommended again. The recommendation accuracy can be improved by continuously optimizing the user portraits and continuously optimizing the weights of various recommendation algorithms.

Referring to fig. 11, fig. 11 is an exemplary diagram of a recommendation flow in a service scenario provided by an embodiment of the present application. As shown in fig. 11, the user tag (corresponding to the user portrait) and the business data tag are acquired and updated respectively, then the matching weights of the user tag and the business data tag are calculated to calculate the score, and then the business data are sorted according to the score so as to be recommended or displayed according to the sorting.

Specifically, the user performs operation behaviors such as riding, consumption, participation activities, transactions and the like at the APP, the data is sent to a user data analysis center, the user data analysis center analyzes the data to perform user portrait reprocessing, after updating user information, the user re-acquires latest tag information, then the background acquires corresponding business data, the tags of the user and the business data tags perform various matching calculation scores, the calculation is performed again according to the weights, finally the score results of the business data are obtained, the business data are ranked according to the scores, the higher the score is, the more the ranking is, and then the corresponding data of the user are returned.

When the user takes the subway every day, the travel data are uploaded to a user travel data analysis center, the user travel data analysis center analyzes the data, and then travel behavior labels of the user are generated and updated according to the data of the user taking the subway and the data of the history taking the subway. And an operator creates an operation commodity on the operation platform and actively marks the commodity or automatically generates an attribute label when the commodity is analyzed.

Illustratively, when the user a sits on the subway from the B site to the C site in the morning, after the travel data of the user is sent up, a travel behavior label of the user is generated: the subway riding user tag, the common station B and the common station C, and then an operator creates a subway month card and breakfast coupons nearby the station C, so that when the engine performs matching pushing through the traveling behavior tag of the user and the attribute tag of the commodity according to the matching policy engine, the user can see the corresponding subway month card and the commodity of the breakfast coupon of the station C on a page.

Referring to fig. 12 to fig. 14, fig. 12 is a schematic flow chart of active recommendation provided by the embodiment of the present application, fig. 13 is a schematic flow chart of passive recommendation provided by the embodiment of the present application, and fig. 14 is a detailed schematic flow chart of passive recommendation provided by the embodiment of the present application. In the embodiment of the application, the APP front end is used as a display and operation layer, and has the main functions of: a. providing a display of content such as commercial activities, advertisements, commodities, parking and the like; b. function buttons providing business operations, such as selecting a business point, pulling down business information, etc.; c. providing a function of viewing personal transaction record information; d. providing a function of viewing personal basic information; e. and carrying out data communication with the back end, and calling a service interface of the back end, wherein the service interface comprises a recommended content list calling interface, a personal information acquisition interface, a recommended content detail acquisition interface, an event uploading interface and the like. The business service is used as an actual business processing layer, and has the main functions of: a. an actual service processing layer, such as processing parking service in passenger parking; b. service triggering generates service event uploading; c. reporting messages on the event of entering and exiting; d. reporting user position information data information; e. reporting the integral deduction message; f. reporting a newly added event triggering message; g. and reporting the event message of insufficient credit of the card. Bff serves as a business aggregation layer, and the main functions are as follows: a. each business service in the business domain is collected and invoked; b. external requests are managed in a unified way. The MQ service serves as a message middleware layer, and the main functions are as follows: a. asynchronous transfer processing of the system; b. service decoupling; c. event stream, bridging business service; d. peak clipping and valley filling. The portrait service is used as a user portrait core processing layer, and has the main functions of: a. carrying out label quantization for passengers and carrying out portrait processing; b. classifying the portraits for the passenger group; c. continuously optimizing the passenger portraits according to the passenger information and the behavior data; d. labeling business information. The recommendation algorithm processing layer with the recommendation service as a core has the main functions of: a. algorithm importing; b. recommending content configuration; c. recommending strategy configuration; d. calculating weight/recommendation algorithm according to the user portrait to obtain corresponding recommendation business points; e. and (5) receiving the event, and adjusting a weight/recommendation algorithm. The management background is used as a business point and basic data editing layer, and has the main functions of: a. providing business point recommended content viewing editing for operators; b. providing information such as algorithm weight for operators; c. providing the capability of editing information such as business points; d. providing a user portrait information viewing capability; e. business policies are configured.

Referring to fig. 15, the embodiment of the present application further provides a content recommendation system 150, including: the user portrait library 1501, the content information library 1502 and the recommendation module 1503 are connected with the recommendation module 1503 by the user portrait library 1501 and the content information library 1502. Wherein:

The user image library 1501 is used for managing and optimizing user images of each user;

the content information library 1502 is used for managing respective content information;

the recommendation module 1503 is configured to execute the recommendation method according to any embodiment of the present application.

The specific implementation of the content recommendation system is basically the same as the specific embodiment of the content recommendation method described above, and will not be described herein.

The embodiment of the application also provides electronic equipment, which comprises a memory and a processor, wherein the memory stores a computer program, and the processor realizes the content recommendation method when executing the computer program. The electronic equipment can be any intelligent terminal including a tablet personal computer, a vehicle-mounted computer and the like.

Referring to fig. 16, fig. 16 illustrates a hardware structure of an electronic device according to another embodiment, the electronic device includes:

The processor 1601 may be implemented by a general-purpose CPU (central processing unit), a microprocessor, an application-specific integrated circuit (ApplicationSpecificIntegratedCircuit, ASIC), or one or more integrated circuits, etc. for executing related programs to implement the technical solution provided by the embodiments of the present application;

memory 1602 may be implemented in the form of read-only memory (ReadOnlyMemory, ROM), static storage, dynamic storage, or random access memory (RandomAccessMemory, RAM). The memory 1602 may store an operating system and other application programs, and when the technical solutions provided in the embodiments of the present disclosure are implemented by software or firmware, relevant program codes are stored in the memory 1602, and the processor 1601 invokes a content recommendation method for executing the embodiments of the present disclosure;

an input/output interface 1603 for implementing information input and output;

The communication interface 1604, configured to implement communication interaction between the present device and other devices, where communication may be implemented through a wired manner (e.g., USB, network cable, etc.), or may be implemented through a wireless manner (e.g., mobile network, WI F I, bluetooth, etc.);

a bus 1605 for transferring information between various components of the device (e.g., processor 1601, memory 1602, input/output interface 1603, and communication interface 1604);

Wherein the processor 1601, the memory 1602, the input/output interface 1603 and the communication interface 1604 enable communication connection with each other inside the device via a bus 1605.

The embodiment of the application also provides a storage medium, which is a computer readable storage medium, and the storage medium stores a computer program, and the computer program realizes the content recommendation method when being executed by a processor.

The memory, as a non-transitory computer readable storage medium, may be used to store non-transitory software programs as well as non-transitory computer executable programs. In addition, the memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one magnetic disk storage device, flash memory device, or other non-transitory solid state storage device. In some embodiments, the memory optionally includes memory remotely located relative to the processor, the remote memory being connectable to the processor through a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.

The embodiments described in the embodiments of the present application are for more clearly describing the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided by the embodiments of the present application, and those skilled in the art can know that, with the evolution of technology and the appearance of new application scenarios, the technical solutions provided by the embodiments of the present application are equally applicable to similar technical problems.

It will be appreciated by persons skilled in the art that the embodiments of the application are not limited by the illustrations, and that more or fewer steps than those shown may be included, or certain steps may be combined, or different steps may be included.

The above described apparatus embodiments are merely illustrative, wherein the units illustrated as separate components may or may not be physically separate, i.e. may be located in one place, or may be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

Those of ordinary skill in the art will appreciate that all or some of the steps of the methods, systems, functional modules/units in the devices disclosed above may be implemented as software, firmware, hardware, and suitable combinations thereof.

The terms "first," "second," "third," "fourth," and the like in the description of the application and in the above figures, if any, are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments of the application described herein may be implemented in sequences other than those illustrated or otherwise described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.

It should be understood that in the present application, "at least one (item)" means one or more, and "a plurality" means two or more. "and/or" for describing the association relationship of the association object, the representation may have three relationships, for example, "a and/or B" may represent: only a, only B and both a and B are present, wherein a, B may be singular or plural. The character "/" generally indicates that the context-dependent object is an "or" relationship. "at least one of" or the like means any combination of these items, including any combination of single item(s) or plural items(s). For example, at least one (one) of a, b or c may represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", wherein a, b, c may be single or plural.

In the several embodiments provided by the present application, it should be understood that the disclosed apparatus and method may be implemented in other manners. For example, the above-described apparatus embodiments are merely illustrative, and for example, the above-described division of units is merely a logical function division, and there may be another division manner in actual implementation, for example, a plurality of units or components may be combined or may be integrated into another system, or some features may be omitted, or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be an indirect coupling or communication connection via some interfaces, devices or units, which may be in electrical, mechanical or other form.

The units described above as separate components may or may not be physically separate, and components shown as units may or may not be physical units, may be located in one place, or may be distributed over a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, each functional unit in the embodiments of the present application may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units may be implemented in hardware or in software functional units.

The integrated units, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application may be embodied in essence or a part contributing to the prior art or all or part of the technical solution in the form of a software product stored in a storage medium, including multiple instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method of the various embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-only Memory (ROM), a random access Memory (Random Access Memory RAM), a magnetic disk, an optical disk, or other various media capable of storing a program.

The preferred embodiments of the present application have been described above with reference to the accompanying drawings, and are not thereby limiting the scope of the claims of the embodiments of the present application. Any modifications, equivalent substitutions and improvements made by those skilled in the art without departing from the scope and spirit of the embodiments of the present application shall fall within the scope of the claims of the embodiments of the present application.

Claims

1. A content recommendation method, the method comprising:

2. The method of claim 1, wherein after pushing the target recommended content to the target user, the method comprises:

3. The method of claim 1, wherein selecting content information to be recommended from a content information base comprises:

4. The method of claim 1, wherein constructing the user representation comprises:

5. The method according to claim 1, wherein the method further comprises:

acquiring user portrait and travel information of the target user;

6. The method according to claim 1, wherein the method further comprises:

determining target recommended content to be recommended from the content information base;

7. The method of claim 1, wherein calculating the matching degree between each recommended content in the content information to be recommended and the user portrait by using a combination of a plurality of recommendation algorithms includes:

8. A content recommendation system, the recommendation system comprising: the system comprises a user portrait library, a content information library and a recommendation module, wherein the user portrait library and the content information library are connected with the recommendation module;

The content information base is used for managing each content information;

the recommendation module is configured to perform the recommendation method of any one of claims 1-7.

9. An electronic device comprising a memory, a processor, a program stored on the memory and executable on the processor, and a data bus for enabling a connected communication between the processor and the memory, the program when executed by the processor implementing the steps of the method according to any of claims 1 to 7.

10. A storage medium, which is a computer-readable storage medium, for computer-readable storage, characterized in that the storage medium stores one or more programs executable by one or more processors to implement the steps of the method of any one of claims 1 to 7.