WO2020135189A1 - 产品推荐方法、产品推荐系统及存储介质 - Google Patents

产品推荐方法、产品推荐系统及存储介质 Download PDF

Info

Publication number
WO2020135189A1
WO2020135189A1 PCT/CN2019/126386 CN2019126386W WO2020135189A1 WO 2020135189 A1 WO2020135189 A1 WO 2020135189A1 CN 2019126386 W CN2019126386 W CN 2019126386W WO 2020135189 A1 WO2020135189 A1 WO 2020135189A1
Authority
WO
WIPO (PCT)
Prior art keywords
attribute
score
product
content
predicted
Prior art date
Application number
PCT/CN2019/126386
Other languages
English (en)
French (fr)
Inventor
徐永泽
赖长明
Original Assignee
深圳Tcl新技术有限公司
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 深圳Tcl新技术有限公司 filed Critical 深圳Tcl新技术有限公司
Priority to EP19904521.2A priority Critical patent/EP3905179A4/en
Priority to US17/258,455 priority patent/US20210295415A1/en
Publication of WO2020135189A1 publication Critical patent/WO2020135189A1/zh

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00Commerce
    • G06Q30/06Buying, selling or leasing transactions
    • G06Q30/0601Electronic shopping [e-shopping]
    • G06Q30/0631Item recommendations
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00Commerce
    • G06Q30/02Marketing; Price estimation or determination; Fundraising
    • G06Q30/0201Market modelling; Market analysis; Collecting market data
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00Commerce
    • G06Q30/02Marketing; Price estimation or determination; Fundraising
    • G06Q30/0201Market modelling; Market analysis; Collecting market data
    • G06Q30/0202Market predictions or forecasting for commercial activities
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00Commerce
    • G06Q30/02Marketing; Price estimation or determination; Fundraising
    • G06Q30/0282Rating or review of business operators or products
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00Commerce
    • G06Q30/06Buying, selling or leasing transactions
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/08Learning methods

Definitions

  • This application relates to the field of communication technology, in particular to a product recommendation method, a product recommendation device, and a storage medium.
  • Websites like Amazon, Netflix, and Spotify all use recommendation systems to recommend products to users.
  • Personalized recommendation systems can bring users considerable convenience when searching for products, thereby enhancing the user experience.
  • Personalized recommendation usually uses collaborative filtering to recommend products to users.
  • the prediction of user interest is based on the analysis of the tastes and preferences of other users in the system, and implicitly infers the "between" "Similarity", the underlying assumption is that when two people have similar tastes, they will have a higher likelihood of maintaining the same attitude towards the product.
  • the existing recommendation system generally regards different products as independent individuals, and extracts and establishes a hidden relationship network between users and users, and between products and products through the user's score (or similar) of different products. , Or will consider the user's location, user rating time information, etc.
  • the recommended target product is a movie.
  • each movie has its own attributes, such as: style, director, actor, language, region, shooting time, awards, etc.
  • attributes such as: style, director, actor, language, region, shooting time, awards, etc.
  • users of the star chasing family will attach great importance to the actor attributes of the movie, and new film lovers will pay more attention to the film's shooting time.
  • the main purpose of this application is to propose a product recommendation method, a product recommendation system, and a readable storage medium, aimed at improving the accuracy of the product recommendation method and enhancing the user experience.
  • This application provides a product recommendation method.
  • the method includes the following steps:
  • the present application also provides a product recommendation system, the system includes: a memory, a processor, and a product recommendation program stored on the memory and runable on the processor, the product When the recommendation program is executed by the processor, the steps of the product recommendation method described above are realized.
  • the present application also provides a storage medium on which a product recommendation program is stored, which implements the steps of the product recommendation method described above when the product recommendation program is executed by the processor.
  • the score of the first attribute content is obtained by scoring multiple products in the same type, the attribute of the product is introduced into the recommendation method, and then the score of the second attribute of the predicted product is calculated according to the score of the first attribute, and Obtain the proportion of each second attribute in the product through the deep neural network model, calculate the score of the predicted product by the score of the second attribute and the proportion of the score of the second attribute, and then recommend the product according to the level of the product score, Considering product attribute information based on the user's rating of the product to achieve product recommendation, not only improves the accuracy of the product recommendation method, but also enhances the user's experience.
  • FIG. 1 is a schematic diagram of a terminal structure of a hardware operating environment involved in an embodiment of the present application
  • FIG. 2 is a schematic flowchart of a first embodiment of a product recommendation method of this application
  • FIG. 3 is a schematic flowchart of a second embodiment of a product recommendation method of this application.
  • FIG. 4 is a schematic flowchart of a third embodiment of a product recommendation method of this application.
  • FIG. 5 is a schematic flowchart of a fourth embodiment of a product recommendation method of this application.
  • FIG. 6 is a schematic flowchart of a fifth embodiment of a product recommendation method of this application.
  • FIG. 7 is a schematic flowchart of a sixth embodiment of a product recommendation method of this application.
  • FIG. 8 is a schematic flowchart of a seventh embodiment of a product recommendation method of this application.
  • FIG. 9 is a schematic flowchart of an eighth embodiment of a product recommendation method of this application.
  • the main solution of the embodiment of the present application is to extract product attribute information, consider the product attribute information based on the user's rating of the product to implement product recommendation, improve the accuracy of the product recommendation method, and enhance the user experience.
  • Existing product recommendation methods simply extract information from users' ratings (or similar) of different products and establish user-to-user, product-to-product implicit relationship networks for product recommendation; After scoring, build a user-product matrix.
  • the rows of this matrix represent a user, and its columns represent a product.
  • the elements in the matrix represent the user's rating (or similar) information about the product. And when the user does not have a rating (or similar) behavior for the product, the element in the matrix is recorded as 0 (zero element), so the user-product matrix is often an extremely sparse matrix.
  • the task of the recommendation system based on the collaborative filtering method is to complete the filling of the zero-element position of the matrix.
  • the prediction matrix of the output of the entire recommendation system is a matrix that is the same size as the original user-product matrix.
  • Each element in the prediction matrix shows the predicted and estimated user’s likes and dislikes of the product.
  • This product recommends The method does not comprehensively consider the attributes of the product to enhance the accuracy of the recommended method.
  • This application obtains the score of the first attribute content by scoring multiple products in the same type, introduces the attribute of the product into the recommendation method, and then calculates the score of the predicted second attribute of the product according to the score of the first attribute, and multiple users
  • the historical scores of multiple products in the same type are used as the input of the collaborative filtering system to obtain the collaborative prediction scores of the predicted products
  • the deep neural network model is used to obtain the score ratio of each second attribute in the product and the collaborative prediction score ratio, according to The prediction score of the second attribute, the ratio of the prediction score of the second attribute, the collaborative prediction score, and the proportion of the collaborative prediction score calculate the scores of the predicted products corresponding to multiple users, and then perform the product according to the level of the product score Recommendation, considering product attribute information based on the user's rating of the product to achieve product recommendation, not only improves the accuracy of the product recommendation method, but also enhances the user's experience.
  • FIG. 1 is a schematic diagram of a terminal structure of a hardware operating environment involved in a solution of an embodiment of the present application.
  • the terminal may be a PC, a smart phone, a tablet computer, an MP4 (Moving Picture Experts Group Audio Layer IV, motion picture expert compression standard audio layer 3) player, a portable computer, and other portable devices with display functions Terminal Equipment.
  • MP4 Motion Picture Experts Group Audio Layer IV, motion picture expert compression standard audio layer 3
  • the terminal may include: a processor 1001, such as a CPU, a network interface 1004, a user interface 1003, a memory 1005, and a communication bus 1002.
  • the communication bus 1002 is configured to implement connection communication between these components.
  • the user interface 1003 may include a display screen (Display), an input unit such as a keyboard (Keyboard), and the optional user interface 1003 may further include a standard wired interface and a wireless interface.
  • the network interface 1004 may optionally include a standard wired interface and a wireless interface (such as a WI-FI interface).
  • the memory 1005 may be a high-speed RAM memory or a non-volatile memory (non-volatile memory), such as a magnetic disk memory.
  • the memory 1005 may optionally be a storage device independent of the foregoing processor 1001.
  • the terminal may also include a camera, RF (Radio Frequency (radio frequency) circuits, sensors, audio circuits, WiFi modules, etc.
  • sensors such as light sensors, motion sensors and other sensors.
  • the light sensor may include an ambient light sensor and a proximity sensor, wherein the ambient light sensor may adjust the brightness of the display screen according to the brightness of the ambient light, and the proximity sensor may turn off the display screen and/or when the mobile terminal moves to the ear Backlight.
  • the gravity acceleration sensor can detect the magnitude of acceleration in various directions (generally three axes), and can detect the magnitude and direction of gravity when at rest, and can be set to applications that recognize the posture of mobile terminals (such as horizontal and vertical screen switching) , Related games, magnetometer posture calibration), vibration recognition related functions (such as pedometer, tap), etc.
  • the mobile terminal can also be configured with other sensors such as gyroscopes, barometers, hygrometers, thermometers, infrared sensors, etc. This will not be repeated here.
  • FIG. 1 does not constitute a limitation on the terminal, and may include more or less components than those illustrated, or combine certain components, or have different component arrangements.
  • the memory 1005 as a computer storage medium may include an operating system, a network communication module, a user interface module, and a product recommendation program.
  • the network interface 1004 is mainly configured to connect to a background server and perform data communication with the background server;
  • the user interface 1003 is mainly configured to connect to a client (user side) and perform data communication with the client;
  • the processor 1001 can be set to call the product recommendation program stored in the memory 1005 and perform the following operations:
  • FIG. 2 is a schematic flowchart of a first embodiment of a method of the present application. The method includes:
  • Step S10 Acquire the scores of the content of the first attribute among the first attributes of the products according to the historical scores of multiple products of the same type by multiple users;
  • the product has the first attribute and the content of the first attribute in the first attribute, for example, when the product is a movie, multiple products in the same type, that is, many Movies, rating multiple movies, movies have first attributes such as style, director, actor, language, region, shooting time, awards, etc., and have the first attribute content in the first attribute, such as style (first attribute ) Has the first attribute content such as documentary style, fusion style, co-occurrence style, the director (first attribute) has the first attribute content such as director A, director B, and director C, and the actor (first attribute) has the actors A, B
  • the first attribute content such as actors, C actors, etc., or when the product is clothes, the clothes have the first attribute such as color, style, size, brand, etc., and the first attribute content in the first attribute, such as color (first attribute) has White, black, blue and other first attribute content
  • style (first attribute) has first attribute content such as elegant, printing, fashion, evening wear, leisure, etc. Brand
  • the user's rating of the first attribute content is obtained, for example:
  • User A's historical rating of movie A is 3 points, and user A's historical rating of movie B is 4 points;
  • User B's historical rating of movie A is 4 points, and user B's historical rating of movie B is 5 points;
  • the first attribute (director) of movie A is director A, and the first attribute (actor) of movie A is actor A and actor B;
  • the first attribute (director) of movie B is director B, and the first attribute (actor) of movie B is actor A and actor C;
  • user A has a score of 3.5 or 7 for actor A, that user A has a score of 3 for actor B, and that user A has a score of 4 for actor C;
  • User B's rating for actor A is 4.5 points or 9 points.
  • User A's rating for actor B is 4 points, and user A's rating for actor C is 5 points;
  • a user-attribute content matrix may be established separately according to the user's rating of the first attribute content.
  • a user-director content matrix may be established based on the above director content rating, based on the above-mentioned actors
  • the content rating establishes the user-actor content matrix, thereby obtaining the user's rating of the director content and the actor content, that is, the first attribute content rating, and the user-first attribute content matrix can be established according to the first attribute content rating
  • Step S20 Obtain the second attribute of the predicted product and the second attribute content of the second attribute, and obtain the user's predicted score for the second attribute of the predicted product according to the score of the first attribute content and the second attribute content;
  • the director of movie C is director A
  • the actors are actor B
  • actor C and the rating of movie C is predicted
  • the predicted products include products not rated by the user and products with historical ratings, and the scores of products that already have historical ratings are re-predicted according to the obtained ratings of the first attribute content.
  • the user-director matrix can be established [4 5], and a matrix of user-actor attributes [4] 5] or
  • Step S30 taking the historical scores of multiple products of the same type by multiple users as the input of the collaborative filtering system to obtain the collaborative prediction score of the predicted product;
  • the collaborative filtering system enters the user's historical score for multiple products in the same type in the collaborative filtering system to obtain the collaborative prediction score of the predicted product, that is, to obtain the predicted score of the attribute of the predicted product itself.
  • the predicted product includes the user's unrated products and
  • the collaborative filtering system obtains a collaborative prediction score for predicted products based on the implicit relationship between multiple products with historical ratings and unrated products and the relationship between users and users. Products, based on the aforementioned relationship between users and users, and the implicit relationship between multiple products with historical ratings and unrated products, re-predict the ratings given to products with historical ratings.
  • the collaborative prediction score shows collaborative filtering The system predicts and estimates how much users like and dislike products.
  • Step S40 Use the score of the first attribute content and the historical score as the input of the deep neural network model to obtain the score ratio of the second attribute and the score of the collaborative prediction score;
  • the scores of the first attribute content (according to the above example, actor A, actor B, actor C, director A, director B) and the scores of movie A and movie B are used as the input of the deep neural network model, Obtain the proportion of the scores of the second attribute such as director and actor and the proportion of the collaborative prediction score (the attribute score of itself).
  • the network structure of the deep neural network model (the depth of the network, the number of neurons used in each layer) can be adjusted according to the structure of the actual data.
  • Step S50 Calculate the scores of the predicted products corresponding to multiple users according to the predicted score of the second attribute, the predicted score ratio of the second attribute, and the collaborative predicted score ratio;
  • the score of each second attribute is multiplied by the score ratio of each second attribute plus the collaborative prediction score multiplied by the collaborative prediction score ratio to obtain the score of the predicted product, for example: suppose that the director’s ratio is obtained through a deep neural network model 30%, the actor accounted for 50%, and the collaborative prediction score accounted for 20%.
  • the user’s collaborative prediction score for the prediction product (Movie C) is [3] 5]
  • [user A’s rating of movie C is: actor attribute rating *50%+ director Attribute score*30%+Collaborative prediction score*20%
  • User B’s score for Movie C is: actor attribute score*50%+director attribute score*30%+collaborative prediction score*20%]
  • the score of D is [3 6]
  • the final prediction matrix is obtained.
  • the scores of the predicted products corresponding to a plurality of the users may also be calculated according to the predicted scores of the second attributes of the products and the weighted average of the collaborative predicted scores.
  • Step S60 Product recommendation is performed for the user according to the level of the predicted product.
  • the user A is recommended to the user A according to the predicted score of the user A for the movie A, movie B, movie C, and movie D.
  • Recommend user B according to the predicted scores of user A for movie A, movie B, movie C, and movie D; or recommend movies within preset score values to users A and B. .
  • the score of the first attribute content is obtained by scoring multiple products in the same type, the attribute of the product is introduced into the recommendation method, and then the score of the second attribute of the predicted product is calculated according to the score of the first attribute.
  • Products and products with historical scores, and the deep neural network model is used to obtain the proportion of each second attribute in the predicted product and the proportion of the collaborative prediction score of the predicted product, through the score of the second attribute, the score of the second attribute.
  • the proportion plus the collaborative prediction score and the proportion of the collaborative prediction score calculate the score of the predicted product, and then recommend the product based on the product score.
  • the product recommendation is not only improved.
  • the accuracy of the recommended method also improves the user experience.
  • FIG. 3 is a schematic flowchart of a second embodiment of the method of the present application. Based on the embodiment shown in FIG. 2 above, step S10 may include:
  • Step S11 According to the historical scores of multiple products of the same type by the user, when only one of the products has the first attribute content, the historical score of the product with the first attribute content is used as the first attribute content score;
  • Step S12 According to the historical scores of multiple products in the same type by the user, when all of the products have the first attribute content, sum the historical scores of the multiple products with the first attribute content as the first A score of attribute content.
  • the historical score of the product with the first attribute content may be used as the score of the first attribute content; the first In the case of attribute content, the historical scores of multiple products with the first attribute content are summed as the score of the first attribute content;
  • user A's historical rating of movie A is 3 points
  • user A's historical rating of movie B is 4 points
  • the first attribute (director) of movie A is director A, and the first attribute (actor) of movie A is actor A and actor B;
  • the first attribute (director) of movie B is director B, and the first attribute (actor) of movie B is actor A and actor C;
  • the user A's rating for actor B is 3 points
  • the user A's rating for actor C is 4 points.
  • FIG. 4 is a schematic flowchart of a third embodiment of a product recommendation method of this application. Based on the above embodiment, step S10 may include:
  • Step S11 According to the historical scores of multiple products in the same type by the user, when only one of the products has the first attribute content, the historical score of the product with the first attribute content is used as the first attribute content score;
  • Step S13 According to the historical scores of a plurality of products of the same type by the user, when all of the products have the first attribute content, the average of the historical scores of the multiple products with the first attribute content is taken as the The score of the first attribute content.
  • the historical score of the product with the first attribute content may be used as the score of the first attribute content; the first In the case of attribute content, the historical scores of multiple products with the first attribute content are averaged as the score of the first attribute content;
  • user A's historical rating of movie A is 3 points
  • user A's historical rating of movie B is 4 points
  • the first attribute (director) of movie A is director A, and the first attribute (actor) of movie A is actor A and actor B;
  • the first attribute (director) of movie B is director B, and the first attribute (actor) of movie B is actor A and actor C;
  • the user A's rating for actor B is 3 points
  • the user A's rating for actor C is 4 points.
  • FIG. 5 is a schematic flowchart of a fourth embodiment of a product recommendation method according to this application. Based on the above embodiment, step S10 may include:
  • Step S14 According to the historical scores of multiple products in the same type by the user, when only one of the products has the first attribute content, the historical score of the product with the first attribute content is normalized as the The score of the first attribute content;
  • Step S15 According to the historical scores of multiple products of the same type by the user, when there are multiple products with the first attribute content, the historical scores of the multiple products with the first attribute content are averaged and then returned The normalization process is used as the score of the first attribute content.
  • the historical score of the product with the first attribute content may be normalized with the highest score made by the user as the first attribute content. Rating; when there are multiple products with the first attribute content, the historical scores of multiple products with the first attribute content are averaged and normalized with the highest score made by the user as the The score of the first attribute content;
  • user A's historical rating of movie A is 3 points
  • user A's historical rating of movie B is 4 points
  • the first attribute (director) of movie A is director A, and the first attribute (actor) of movie A is actor A and actor B;
  • the first attribute (director) of movie B is director B, and the first attribute (actor) of movie B is actor A and actor C;
  • the embodiment it is not limited to the maximum value method, the average value method and the normalization processing method in the above embodiment to obtain the score of the first attribute content, and the minimum value method, weighted average method and random sampling may also be used Other methods obtain the score of the first attribute content.
  • FIG. 6 is a schematic flowchart of a fifth embodiment of a product recommendation method according to this application. Based on the above embodiment, step S20 may include:
  • Step S21 Obtain the second attribute of the predicted product and the second attribute content in the second attribute.
  • the score of the first attribute content corresponding to the second attribute content is taken as the first The score of the second attribute;
  • Step S22 Obtain the second attribute of the predicted product and the second attribute content in the second attribute.
  • the highest score in the first attribute content corresponding to the second attribute content is taken The score as the second attribute.
  • the score of the first attribute content corresponding to the second attribute content is taken as the score of the second attribute; when there are multiple second attribute contents in the second attribute, the corresponding The highest score in the first attribute content of the second attribute content is taken as the score of the second attribute.
  • step S20 may include:
  • Step S21 Obtain the second attribute of the predicted product and the second attribute content in the second attribute.
  • the score of the first attribute content corresponding to the second attribute content is taken as the first The score of the second attribute;
  • Step S23 Obtain the second attribute of the predicted product and the second attribute content of the second attribute.
  • the average of the scores of the first attribute content corresponding to the second attribute content is taken The value serves as the score for the second attribute.
  • the score of the first attribute content corresponding to the second attribute content is taken as the score of the second attribute; or when there are multiple second attribute content in the second attribute, The average value of the scores in the first attribute content corresponding to the second attribute content is used as the score of the second attribute.
  • FIG. 8 is a schematic flowchart of a seventh embodiment of a product recommendation method according to this application. Based on the above embodiment, step S60 may include:
  • Step S61 Recommend a predicted product greater than a preset recommended score value for the user according to the grade of the predicted product.
  • a preset recommendation score value and when the score of the predicted product is greater than the preset recommendation score value, a recommendation is made. Assuming that the preset recommended rating value is 2 points, and there are 50 users A who have scored more than 2 points for the predicted product, the 50 products are recommended to user A. There are 30 users B who have rated the predicted products greater than 2 points, and then recommend these 30 products to user B.
  • FIG. 9 is a schematic flowchart of an eighth embodiment of a product recommendation method according to this application. Based on the above embodiment, step S60 may include:
  • step S62 a preset recommended number of predicted products are recommended for the user according to the grade of the predicted product.
  • Preset recommendation number based on the predicted product's score, take the preset number of predicted products for recommendation. Assuming that the preset recommended number is 20, then based on the predicted user A's rating of the predicted product, take the top 20 products Recommend to user A.
  • This application also provides a product recommendation system.
  • the product recommendation system of the present application includes: a memory, a processor, and a product recommendation program stored on the memory and executable on the processor.
  • the product recommendation program is implemented by the processor to implement the product described above Recommended method steps.
  • the application also provides a storage medium.
  • a product recommendation program is stored on the storage medium of the present application, and when the processor is executed by the processor, the steps of the product recommendation method described above are implemented.

Landscapes

  • Business, Economics & Management (AREA)
  • Engineering & Computer Science (AREA)
  • Finance (AREA)
  • Accounting & Taxation (AREA)
  • Development Economics (AREA)
  • Strategic Management (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • General Physics & Mathematics (AREA)
  • Economics (AREA)
  • Marketing (AREA)
  • General Business, Economics & Management (AREA)
  • Game Theory and Decision Science (AREA)
  • Data Mining & Analysis (AREA)
  • Artificial Intelligence (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Evolutionary Computation (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

一种产品推荐方法、一种产品推荐系统和一种存储介质,该方法包括:根据多个用户对同类型中多个产品的历史评分,获取所述产品的第一属性内容的评分(S10);根据第一属性内容的评分以及所述第二属性内容获得用户对预测产品的第二属性的预测评分(S20);将多个产品的历史评分作为协同滤波系统的输入,获取预测产品的协同预测评分(S30);获取第二属性的预测评分占比以及协同预测评分占比(S40);根据第二属性的预测评分、所述第二属性的预测评分占比、协同预测评分以及协同预测评分占比计算多个所述用户对应的所述预测产品的评分(S50);根据预测产品的评分高低针对所述用户进行产品推荐(S60)。上述方法能够提高产品推荐方法的准确性,提升用户体验。

Description

产品推荐方法、产品推荐系统及存储介质
本申请要求于2018年12月29日提交中国专利局、申请号为201811654674.8、发明名称为“产品推荐方法、产品推荐系统及存储介质”的中国专利申请的优先权,其全部内容通过引用结合在申请中。
技术领域
本申请涉及通信技术领域,尤其涉及产品推荐方法、产品推荐装置及存储介质。
背景技术
像亚马逊、Netflix和Spotify这样的网站都会使用推荐系统给用户推荐产品,个性化推荐系统能够为用户在产品搜索时带来相当的便利,以此提升用户体验。个性化推荐通常使用协同滤波的方法向用户推荐产品,在这种方法中,用户兴趣的预测是基于对系统中其他用户的品味和偏好的分析,并隐含地推断出两者之间的“相似性”,潜在的假设是当两个人有相似的品味,他们将有更高的可能性对产品保有相同的态度。
如现有的推荐系统一般会把不同的产品视作独立的个体,通过用户对不同产品的评分(或对应的类似)信息提取并建立出用户与用户,产品与产品之间隐性的关系网络,或者会考虑用户的位置,用户评分的时间信息等。
然而在实际应用场景中,除了用户对产品的评分(或类似的)记录外,还可以获取大量其它信息,特别是关于产品的属性信息,以电影推荐应用为例,推荐的目标产品是电影,每部电影除了有不同用户评分的记录外还拥有自身的属性,如:风格、导演、演员、语言、地区、拍摄时间、获奖情况等。当用户选择观看电影,很多时候都是在这些属性的基础上进行选择的。对不同属性的选择也反应了用户的类型,例如:追星族型的用户就会对电影的演员属性非常重视,而新片爱好者就更加关注电影的拍摄时间。
单纯地使用用户对不同产品的评分得到隐藏关系网的推荐方法具有一定的效果。但是忽略产品本身的固有属性的差别,不考虑用户对产品显示属性的参考无疑是信息的浪费。反之,有效地参考这些信息结合协同滤波方法本身的优秀能力势必可以强化推荐方法的准确性。
技术问题
本申请的主要目的在于提出一种产品推荐方法、产品推荐系统及可读存储介质,旨在提高产品推荐方法的准确性,提升用户体验。
技术解决方案
本申请提供一种产品推荐方法,所述方法包括如下步骤:
根据多个用户对同类型中多个产品的历史评分,获取多个所述产品的第一属性中第一属性内容的评分;
获取预测产品的第二属性以及第二属性中的第二属性内容,根据第一属性内容的评分以及所述第二属性内容获得用户对预测产品的第二属性的预测评分;
将多个用户对同类型中多个产品的历史评分作为协同滤波系统的输入,获取预测产品的协同预测评分;
将第一属性内容的评分以及所述历史评分作为深度神经网络模型的输入,获取第二属性的预测评分占比以及协同预测评分占比;
根据第二属性的预测评分、所述第二属性的预测评分占比、协同预测评分以及协同预测评分占比计算多个所述用户对应的所述预测产品的评分;
根据预测产品的评分高低针对所述用户进行产品推荐。
此外,为实现上述目的,本申请还提供一种产品推荐系统,所述系统包括:存储器、处理器及存储在所述存储器上并可在所述处理器上运行的产品推荐程序,所述产品推荐程序被所述处理器执行时实现如上所述的产品推荐方法的步骤。
此外,为实现上述目的,本申请还提供一种存储介质,所述存储介质上存储有产品推荐程序,所述产品推荐程序被处理器执行时实现如上所述的产品推荐方法的步骤。
有益效果
本申请实施例中通过在同类型中多个产品的评分获得第一属性内容的评分,将产品的属性引入推荐方法中,再根据第一属性的评分计算预测产品的第二属性的评分,且通过深度神经网络模型获取获得产品中各第二属性的占比,通过第二属性的评分以及所述第二属性的评分占比计算出预测产品的评分,再根据产品评分的高低进行产品推荐,在用户对产品的评分基础上考虑产品属性信息实现产品推荐,不仅提高了产品推荐方法的准确性,还提升了用户的体验。
附图说明
图1是本申请实施例方案涉及的硬件运行环境的终端结构示意图;
图2为本申请产品推荐方法第一实施例的流程示意图;
图3为本申请产品推荐方法第二实施例的流程示意图;
图4为本申请产品推荐方法第三实施例的流程示意图;
图5为本申请产品推荐方法第四实施例的流程示意图;
图6为本申请产品推荐方法第五实施例的流程示意图;
图7为本申请产品推荐方法第六实施例的流程示意图;
图8为本申请产品推荐方法第七实施例的流程示意图;
图9为本申请产品推荐方法第八实施例的流程示意图。
本申请目的的实现、功能特点及优点将结合实施例,参照附图做进一步说明。
本发明的实施方式
应当理解,此处所描述的具体实施例仅仅用以解释本申请,并不设置为限定本申请。
本申请实施例的主要解决方案是:提取产品属性信息,在用户对产品的评分基础上考虑产品属性信息实现产品推荐,提高产品推荐方法的准确性,提升用户的体验。
现有的产品推荐方法单纯通过用户对不同产品的评分(或对应的类似)信息提取并建立出用户与用户,产品与产品之间隐性的关系网络进行产品推荐;如在获取用户对产品的评分后,建立一个用户-产品的矩阵。这个矩阵的行代表一个用户,它的列代表一个产品,矩阵中的元素表示用户对产品的评分(或对应的类似)信息。且当用户对产品不存在评分(或对应的类似)行为时,将矩阵中的元素记为0(零元素),因此用户-产品的矩阵往往会是一个极度稀疏的矩阵。基于协同滤波方法的推荐系统的任务就是完成对矩阵零元素位置的填充,填充的准则则来源于矩阵非零元素显示的用户-用户,产品-产品之间的隐性关系。所以,整个推荐系统的输出的预测矩阵,是一个和原用户-产品矩阵大小一致的矩阵,预测矩阵中的每个元素显示了预测、估计的用户对产品的喜恶程度,这种产品推荐的方法未综合考虑产品的属性以强化推荐方法的准确性。
本申请通过在同类型中多个产品的评分得到第一属性内容的评分,将产品的属性引入推荐方法中,再根据第一属性的评分计算预测产品的第二属性的评分,将多个用户对同类型中多个产品的历史评分作为协同滤波系统的输入,获取预测产品的协同预测评分,且通过深度神经网络模型获得产品中各第二属性的评分占比以及协同预测评分占比,根据第二属性的预测评分、所述第二属性的预测评分占比、协同预测评分以及协同预测评分占比计算多个所述用户对应的所述预测产品的评分,再根据产品评分的高低进行产品推荐,在用户对产品的评分基础上考虑产品属性信息实现产品推荐,不仅提高了产品推荐方法的准确性,还提升了用户的体验。
如图1所示,图1是本申请实施例方案涉及的硬件运行环境的终端结构示意图。
本申请实施例终端可以是PC,也可以是智能手机、平板电脑、MP4(Moving Picture Experts Group Audio Layer IV,动态影像专家压缩标准音频层面3)播放器、便携计算机等具有显示功能的可移动式终端设备。
如图1所示,该终端可以包括:处理器1001,例如CPU,网络接口1004,用户接口1003,存储器1005,通信总线1002。其中,通信总线1002设置为实现这些组件之间的连接通信。用户接口1003可以包括显示屏(Display)、输入单元比如键盘(Keyboard),可选用户接口1003还可以包括标准的有线接口、无线接口。网络接口1004可选的可以包括标准的有线接口、无线接口(如WI-FI接口)。存储器1005可以是高速RAM存储器,也可以是稳定的存储器(non-volatile memory),例如磁盘存储器。存储器1005可选的还可以是独立于前述处理器1001的存储装置。
可选地,终端还可以包括摄像头、RF(Radio Frequency,射频)电路,传感器、音频电路、WiFi模块等等。其中,传感器比如光传感器、运动传感器以及其他传感器。具体地,光传感器可包括环境光传感器及接近传感器,其中,环境光传感器可根据环境光线的明暗来调节显示屏的亮度,接近传感器可在移动终端移动到耳边时,关闭显示屏和/或背光。作为运动传感器的一种,重力加速度传感器可检测各个方向上(一般为三轴)加速度的大小,静止时可检测出重力的大小及方向,可设置为识别移动终端姿态的应用(比如横竖屏切换、相关游戏、磁力计姿态校准)、振动识别相关功能(比如计步器、敲击)等;当然,移动终端还可配置陀螺仪、气压计、湿度计、温度计、红外线传感器等其他传感器,在此不再赘述。
本领域技术人员可以理解,图1中示出的终端结构并不构成对终端的限定,可以包括比图示更多或更少的部件,或者组合某些部件,或者不同的部件布置。
如图1所示,作为一种计算机存储介质的存储器1005中可以包括操作系统、网络通信模块、用户接口模块以及产品推荐程序。
在图1所示的终端中,网络接口1004主要设置为连接后台服务器,与后台服务器进行数据通信;用户接口1003主要设置为连接客户端(用户端),与客户端进行数据通信;而处理器1001可以设置为调用存储器1005中存储的产品推荐程序,并执行以下操作:
根据多个用户对同类型中多个产品的历史评分,获取多个所述产品的第一属性中第一属性内容的评分;
获取预测产品的第二属性以及第二属性中的第二属性内容,根据第一属性内容的评分以及所述第二属性内容获得用户对预测产品的第二属性的预测评分;
将多个用户对同类型中多个产品的历史评分作为协同滤波系统的输入,获取预测产品的协同预测评分;
将第一属性内容的评分以及所述历史评分作为深度神经网络模型的输入,获取第二属性的预测评分占比以及协同预测评分占比;
根据第二属性的预测评分、所述第二属性的预测评分占比、协同预测评分以及协同预测评分占比计算多个所述用户对应的所述预测产品的评分;
根据预测产品的评分高低针对所述用户进行产品推荐。
其中,在处理器1001调用存储器1005中存储的产品推荐程序被执行时所实现的方法可参照本申请产品推荐方法各个实施例,此处不再赘述。
基于上述硬件结构,提出本申请方法实施例。
参照图2,图2为本申请方法第一实施例的流程示意图,所述方法包括:
步骤S10,根据多个用户对同类型中多个产品的历史评分,获取多个所述产品的第一属性中第一属性内容的评分;
获取用户对同类型多个产品的历史评分,所述产品具有第一属性以及具有所述第一属性中的第一属性内容,例如,产品是电影时,同类型中的多个产品,即多部电影,对多部电影的评分,电影有风格、导演、演员、语言、地区、拍摄时间、获奖情况等第一属性,且有第一属性中的第一属性内容,如风格(第一属性)具有纪实风格、溶合风格、共现风格等第一属性内容,导演(第一属性)具有A导演、B导演、C导演等第一属性内容,演员(第一属性)具有A演员、B演员、C演员等第一属性内容,或者产品是衣服时,衣服有颜色、款式、尺寸、品牌等第一属性,且有第一属性中的第一属性内容,如颜色(第一属性)具有白色、黑色、蓝色等第一属性内容,款式(第一属性)具有典雅、印花、时尚、晚装、休闲等第一属性内容,品牌(第一属性)具有A品牌、B品牌、C品牌等第一属性内容,本领域的技术人员应该可以理解,这里只是设置为举例说明,不应理解为对本申请的限制。
可以理解地,根据用户对同类型产品中的多个的历史评分,获取用户对第一属性内容的评分,例如:
用户A对电影A的历史评分为3分,用户A对电影B的历史评分为4分;
用户B对电影A的历史评分为4分,用户B对电影B的历史评分为5分;
电影A的第一属性(导演)为导演A,电影A的第一属性(演员)为演员A、演员B;
电影B的第一属性(导演)为导演B,电影B的第一属性(演员)为演员A、演员C;
可得到用户A对导演A的评分为3分,用户A对导演B的评分为4分,
用户B对导演A的评分为4分,用户B对导演B的评分为5分,
可得到用户A对演员A的评分为3.5分或者7分,可得到用户A对演员B的评分为3分,用户A对演员C的评分为4分;
用户B对演员A的评分为4.5分或者9分,可得到用户A对演员B的评分为4分,用户A对演员C的评分为5分;
可以理解地,为方便产品推荐系统运算,可根据用户对第一属性内容的评分分别建立用户-属性内容矩阵,承上举例,可根据上述导演内容的评分建立用户-导演内容矩阵,根据上述演员内容的评分建立用户-演员内容矩阵,从而得到了用户对导演内容、演员内容的评分,即第一属性内容的评分,可根据第一属性内容的评分分别建立用户-第一属性内容的矩阵, 本领域的技术人员应该可以理解,这里只是设置为举例说明,不应理解为对本申请的限制。
步骤S20,获取预测产品的第二属性以及第二属性中的第二属性内容,根据第一属性内容的评分以及所述第二属性内容获得用户对预测产品的第二属性的预测评分;
获取预测产品的第二属性以及第二属性中的第二属性内容,根据所述第一属性内容的评分以及所述第二属性内容获得预测产品的第二属性的评分;
例如,电影C的导演是导演A,演员是演员B、演员C,对电影C的评分进行预测;
则预测产品电影C的第二属性导演及第二属性导演的第二属性内容导演B,预测产品电影C的第二属性演员及第二属性演员的第二属性内容为演员B、演员C;
则可根据所述第一属性内容的评分,承上所述举例,用户A和B对导演B(第二属性内容)的评分分别为4分、5分,预测产品(电影C)中的第二属性(导演)中只具有一个第二属性内容(导演B),则用户A和B对电影C中的导演的评分分别为4分、5分;
预测产品(电影C)中的第二属性(演员)中具有多个(2个)第二属性内容(演员B、演员C)时;
用户A对演员B评分为3分,用户A对演员C评分为4分;则得到用户A对预测产品(电影C)的第二属性(演员)的评分为4分(取最大值方法获得第二属性的评分)或者3.5分(取平均值方法获得第二属性的评分)。
用户B对演员B评分为4分,用户B对演员C评分为5分,则得到用户B对预测产品(电影C)的第二属性(演员)的评分为5分(取最大值方法获得第二属性的评分)或者4.5分(取平均值方法获得第二属性的评分)。
在实施例中,预测产品包括用户未评分的产品和具有历史评分的产品,根据获得的第一属性内容的评分重新预测已经具有历史评分的产品的评分。
可建立用户-导演的矩阵[4  5],以及用户-演员属性的矩阵[4  5]或者
[3.5  4.5]2个属性矩阵以进行下一步的计算,本领域的技术人员应该可以理解,这里只是设置为举例说明,不应理解为对本申请的限制。
步骤S30,将多个用户对同类型中多个产品的历史评分作为协同滤波系统的输入,获取预测产品的协同预测评分;
可以理解地,在协同滤波系统中输入用户对同类型中的多个产品的历史评分,获得预测产品的协同预测评分,即获得预测产品本身属性的预测评分,预测产品包括用户未评分的产品和具有历史评分的产品,协同滤波系统根据输入有历史评分的多个产品与未评分产品之间的隐性关系,以及用户与用户之间的关系,获得预测产品的协同预测评分,对具有历史评分的产品,则根据前述用户与用户之间的关系,以及有历史评分的多个产品与未评分产品之间的隐性关系重新预测给予具有历史评分的产品的评分,协同预测评分显示了协同滤波系统预测、估计的用户对产品的喜恶程度。
步骤S40,将第一属性内容的评分以及所述历史评分作为深度神经网络模型的输入,获取第二属性的评分占比以及协同预测评分占比;
可以理解地,例如,将第一属性内容(承上举例,演员A、演员B、演员C、导演A、导演B)的评分以及对电影A、电影B的评分作为深度神经网络模型的输入,获取导演、演员等第二属性的评分占比以及协同预测评分(本身属性评分)的占比。深度神经网络模型的网络结构(网络的深度、每层使用的神经元个数)可根据实际数据的结构进行调节。
步骤S50,根据第二属性的预测评分、所述第二属性的预测评分占比、协同预测评分占比计算多个所述用户对应的所述预测产品的评分;
将各第二属性的评分乘以各第二属性的评分占比加上协同预测评分乘以协同预测评分占比得到所述预测产品的评分,例如:假设通过深度神经网络模型获得导演的占比为30%,演员的占比为50%,协同预测评分占比20%,承上举例,用户对预测产品(电影C)的演员评分为[4  5],用户对预测产品(电影C)的导演评分为[4  5],用户对预测产品(电影C)的协同预测评分为[3  5]则用户电影C的评分为[4  5]*50%+[4  5]*30%+[3  5]*20%=[3.8  5],也可以理解为矩阵[用户A对电影C的评分为:演员属性评分*50%+导演属性评分*30%+协同预测评分*20%  用户B对电影C的评分为:演员属性评分*50%+导演属性评分*30%+协同预测评分*20%] ,假设用户对预测的产品电影D的评分为[3  6],得到最终预测矩阵。
在实施例中,还可以根据产品的各第二属性的预测评分以及协同预测评分的加权平均值计算多个所述用户对应的所述预测产品的评分。
步骤S60,根据预测产品的评分高低针对所述用户进行产品推荐。
针对用户通过预测矩阵中评分高低进行产品的推荐,承上举例,根据用户A对电影A,电影B、电影C,电影D的预测评分高低对用户A进行推荐。根据用户B对电影A,电影B、电影C,电影D的预测评分高低对用户B进行推荐;或者向用户A、B推荐预设评分值内的电影。。
通过在同类型中多个产品的评分得到第一属性内容的评分,将产品的属性引入推荐方法中,再根据第一属性的评分计算预测产品的第二属性的评分,预测产品包括用户未评分的产品和具有历史评分的产品,且通过深度神经网络模型获取预测产品中各第二属性的占比以及预测产品的协同预测评分占比,通过第二属性的评分、所述第二属性的评分占比加上协同预测评分以及协同预测评分占比计算出预测产品的评分,再根据产品评分的高低进行产品推荐,在用户对产品的评分基础上考虑产品属性信息实现产品推荐,不仅提高了产品推荐方法的准确性,还提升了用户的体验。
进一步地,参照图3,图3为本申请方法第二实施例的流程示意图。基于上述图2所示的实施例,步骤S10可以包括:
步骤S11,根据用户对同类型中多个产品的历史评分,在只有一个所述产品中具有第一属性内容时,将具有第一属性内容的该产品的历史评分作为所述第一属性内容的评分;
步骤S12,根据用户对同类型中多个产品的历史评分,在多个所述产品中均具有第一属性内容时,将具有第一属性内容的多个产品的历史评分求和作为所述第一属性内容的评分。
可以理解地,在只有一个产品中具有第一属性内容时,可将具有第一属性内容的该产品的历史评分作为所述第一属性内容的评分;在有多个所述产品中具有第一属性内容时,将具有第一属性内容的多个产品的历史评分求和作为所述第一属性内容的评分;
例如,用户A对电影A的历史评分为3分,用户A对电影B的历史评分为4分;
电影A的第一属性(导演)为导演A,电影A的第一属性(演员)为演员A、演员B;
电影B的第一属性(导演)为导演B,电影B的第一属性(演员)为演员A、演员C;
则用户A对导演A的评分为3分,用户A对导演B的评分为4分,
而用户A对演员A的评分为3+4=7分,用户A对演员B的评分为3分,用户A对演员C的评分为4分。
进一步地,参照图4,图4为本申请产品推荐方法第三实施例的流程示意图。基于上述的实施例,步骤S10可以包括:
步骤S11,根据用户对同类型中多个产品的历史评分,在只有一个所述产品中具有第一属性内容时,将具有第一属性内容的该产品的历史评分作为所述第一属性内容的评分;
步骤S13,根据用户对同类型中多个产品的历史评分,在多个所述产品中均具有第一属性内容时,将具有第一属性内容的多个产品的历史评分取平均值作为所述第一属性内容的评分。
可以理解地,在只有一个产品中具有第一属性内容时,可将具有第一属性内容的该产品的历史评分作为所述第一属性内容的评分;在有多个所述产品中具有第一属性内容时,将具有第一属性内容的多个产品的历史评分取平均值作为所述第一属性内容的评分;
例如用户A对电影A的历史评分为3分,用户A对电影B的历史评分为4分;
电影A的第一属性(导演)为导演A,电影A的第一属性(演员)为演员A、演员B;
电影B的第一属性(导演)为导演B,电影B的第一属性(演员)为演员A、演员C;
则用户A对导演A的评分为3分,用户A对导演B的评分为4分,
而用户A对演员A的评分为(3+4)/2=3.5分,用户A对演员B的评分为3分,用户A对演员C的评分为4分。
进一步地,参照图5,图5为本申请产品推荐方法第四实施例的流程示意图。基于上述的实施例,步骤S10可以包括:
步骤S14,根据用户对同类型中多个产品的历史评分,在只有一个所述产品中具有第一属性内容时,将具有第一属性内容的该产品的历史评分进行归一化处理作为所述第一属性内容的评分;
步骤S15,根据用户对同类型中多个产品的历史评分,在有多个所述产品中具有第一属性内容时,将具有第一属性内容的多个产品的历史评分取平均值后进行归一化处理作为所述第一属性内容的评分。
可以理解地,在只有一个产品中具有第一属性内容时,可将具有第一属性内容的该产品的历史评分以该用户作出的最高评分进行归一化处理后作为所述第一属性内容的评分;在有多个所述产品中具有第一属性内容时,将具有第一属性内容的多个产品的历史评分取平均值后以该用户作出的最高评分进行归一化处理后作为所述第一属性内容的评分;
例如用户A对电影A的历史评分为3分,用户A对电影B的历史评分为4分;
电影A的第一属性(导演)为导演A,电影A的第一属性(演员)为演员A、演员B;
电影B的第一属性(导演)为导演B,电影B的第一属性(演员)为演员A、演员C;
则用户A对导演A的评分为3/4分,用户A对导演B的评分为4/4=1分,
而用户A对演员A的评分为(3+4)/2/4=0.875分,用户A对演员B的评分为3/4分,用户A对演员C的评分为4/4=1分。
在实施例中,不限于上述实施例中取最大值方法、取平均值方法以及归一化处理方法获得第一属性内容的评分,还可以采用取最小值方法、加权平均值方法以及随机采样等其他方法获得第一属性内容的评分。
进一步地,参照图6,图6为本申请产品推荐方法第五实施例的流程示意图。基于上述的实施例,步骤S20可以包括:
步骤S21,获取预测产品的第二属性以及第二属性中的第二属性内容,在第二属性中只具有一个第二属性内容时,将对应第二属性内容的第一属性内容的评分作为第二属性的评分;
步骤S22,获取预测产品的第二属性以及第二属性中的第二属性内容,在第二属性中具有多个第二属性内容时,取对应第二属性内容的第一属性内容中的最高评分作为第二属性的评分。
在第二属性中只具有一个第二属性内容时,将对应第二属性内容的第一属性内容的评分作为第二属性的评分;在第二属性中具有多个第二属性内容时,取对应第二属性内容的第一属性内容中的最高评分作为第二属性的评分。
例如,获取预测产品(电影C)的第二属性(导演)及导演的第二属性内容(导演B),获取预测产品(电影C)的第二属性(演员)及演员的第二属性内容(演员B、演员C);
则可根据所述第一属性内容的评分,承上所述举例,用户A和B对导演B(第二属性内容)的评分分别为4分、5分,预测产品(电影C)中的第二属性(导演)中只具有一个第二属性内容(导演B),则用户A和B对电影C中的导演的评分分别为4分、5分;
预测产品(电影C)中的第二属性(演员)中具有多个(2个)第二属性内容(演员B、演员C)时;
用户A对演员B评分为3分,用户A对演员C评分为4分;则得到用户A对预测产品(电影C)的第二属性(演员)的评分为4分。
用户B对演员B评分为4分,用户B对演员C评分为5分,则得到用户B对预测产品(电影C)的第二属性(演员)的评分为5分。
进一步地,参照图7,图7为本申请产品推荐方法第六实施例的流程示意图。基于上述的实施例,步骤S20可以包括:
步骤S21,获取预测产品的第二属性以及第二属性中的第二属性内容,在第二属性中只具有一个第二属性内容时,将对应第二属性内容的第一属性内容的评分作为第二属性的评分;
步骤S23,获取预测产品的第二属性以及第二属性中的第二属性内容,在第二属性中具有多个第二属性内容时,取对应第二属性内容的第一属性内容的评分的平均值作为第二属性的评分。
在第二属性中只具有一个第二属性内容时,将对应第二属性内容的第一属性内容的评分作为第二属性的评分;或者在第二属性中具有多个第二属性内容时,取对应第二属性内容的第一属性内容中的评分的平均值作为第二属性的评分。
例如,获取预测产品(电影C)的第二属性(导演)及导演的第二属性内容(导演B),获取预测产品(电影C)的第二属性(演员)及演员的第二属性内容(演员B、演员C);
则可根据所述第一属性内容的评分,承上所述举例,用户A和B对导演B(第二属性内容)的评分分别为4分、5分,预测产品(电影C)中的第二属性(导演)中只具有一个第二属性内容(导演B),则用户A和B对电影C中的导演的评分分别为4分、5分;
预测产品(电影C)中的第二属性(演员)中具有多个(2个)第二属性内容(演员B、演员C)时;
用户A对演员B评分为3分,用户A对演员C评分为4分;则得到用户A对预测产品(电影C)的第二属性(演员)的评分为(3+4)/2=3.5分。
用户B对演员B评分为4分,用户B对演员C评分为5分,则得到用户B对预测产品(电影C)的第二属性(演员)的评分为(4+5)/2=4.5分。
进一步地,参照图8,图8为本申请产品推荐方法第七实施例的流程示意图。基于上述的实施例,步骤S60可以包括:
步骤S61,根据预测产品的评分高低针对所述用户推荐大于预设推荐评分值的预测产品。
预设推荐评分值,所述预测产品的评分大于预设的推荐评分值时,则进行推荐。假设预设推荐评分值为2分,用户A对预测产品的评分大于2分的有50个,则对用户A推荐这50个产品。用户B对预测产品的评分大于2分的有30个,则对用户B推荐这30个产品。
进一步地,参照图9,图9为本申请产品推荐方法第八实施例的流程示意图。基于上述的实施例,步骤S60可以包括:
步骤S62,根据预测产品的评分高低针对所述用户推荐预设推荐个数的预测产品。
预设推荐个数,根据预测产品的评分高低取前预设个数预测产品进行推荐,假设预设推荐个数为20个,则根据预测用户A对预测产品的评分高低,取前20个产品向用户A进行推荐。
本申请还提供一种产品推荐系统。
本申请产品推荐系统包括:存储器、处理器及存储在所述存储器上并可在所述处理器上运行的产品推荐程序,所述产品推荐程序被所述处理器执行时实现如上所述的产品推荐方法的步骤。
其中,在所述处理器上运行的产品推荐程序被执行时所实现的方法可参照本申请产品推荐方法各个实施例,此处不再赘述。
本申请还提供一种存储介质。
本申请存储介质上存储有产品推荐程序,所述产品推荐程序被处理器执行时实现如上所述的产品推荐方法的步骤。
其中,在所述处理器上运行的产品推荐程序被执行时所实现的方法可参照本申请产品推荐方法各个实施例,此处不再赘述。
以上仅为本申请的优选实施例,并非因此限制本申请的专利范围,凡是利用本申请说明书及附图内容所作的等效结构或等效流程变换,或直接或间接运用在其他相关的技术领域,均同理包括在本申请的专利保护范围内。

Claims (20)

  1. 一种产品推荐方法,其中,所述方法包括如下步骤:
    根据多个用户对同类型中多个产品的历史评分,获取多个所述产品的第一属性中第一属性内容的评分;
    获取预测产品的第二属性以及第二属性中的第二属性内容,根据第一属性内容的评分以及所述第二属性内容获得用户对预测产品的第二属性的预测评分;
    将多个用户对同类型中多个产品的历史评分作为协同滤波系统的输入,获取预测产品的协同预测评分;
    将第一属性内容的评分以及所述历史评分作为深度神经网络模型的输入,获取第二属性的预测评分占比以及协同预测评分占比,以及;
    根据第二属性的预测评分、所述第二属性的预测评分占比、协同预测评分以及协同预测评分占比计算多个所述用户对应的所述预测产品的评分;
    根据预测产品的评分高低针对所述用户进行产品推荐。
  2. 如权利要求1所述的产品推荐方法,其中,所述根据多个用户对同类型中多个产品的历史评分,获取多个所述产品的第一属性中第一属性内容的评分的步骤包括:
    根据用户对同类型中多个产品的历史评分,在只有一个所述产品中具有第一属性内容时,将具有第一属性内容的该产品的历史评分作为所述第一属性内容的评分,以及;
    根据用户对同类型中多个产品的历史评分,在多个所述产品中均具有第一属性内容时,将具有第一属性内容的多个产品的历史评分求和作为所述第一属性内容的评分。
  3. 如权利要求1所述的产品推荐方法,其中,所述根据多个用户对同类型中多个产品的历史评分,获取多个所述产品的第一属性中第一属性内容的评分的步骤包括:
    根据用户对同类型中多个产品的历史评分,在只有一个所述产品中具有第一属性内容时,将具有第一属性内容的该产品的历史评分作为所述第一属性内容的评分,以及;
    根据用户对同类型中多个产品的历史评分,在多个所述产品中均具有第一属性内容时,将具有第一属性内容的多个产品的历史评分取平均值作为所述第一属性内容的评分。
  4. 如权利要求1所述的产品推荐方法,其中,所述根据多个用户对同类型中多个产品的历史评分,获取多个所述产品的第一属性中第一属性内容的评分的步骤包括:
    根据用户对同类型中多个产品的历史评分,在只有一个所述产品中具有第一属性内容时,将具有第一属性内容的该产品的历史评分进行归一化处理作为所述第一属性内容的评分,以及;
    根据用户对同类型中多个产品的历史评分,在多个所述产品中均具有第一属性内容时,将具有第一属性内容的多个产品的历史评分取平均值后进行归一化处理作为所述第一属性内容的评分。
  5. 如权利要求1所述的产品推荐方法,其中,所述获取预测产品的第二属性以及第二属性中的第二属性内容,根据第一属性内容的评分以及所述第二属性内容获得用户对预测产品的第二属性的预测评分的步骤包括:
    获取预测产品的第二属性以及第二属性中的第二属性内容,在第二属性中只具有一个第二属性内容时,将对应第二属性内容的第一属性内容的评分作为第二属性的评分,以及;
    获取预测产品的第二属性以及第二属性中的第二属性内容,在第二属性中具有多个第二属性内容时,取对应第二属性内容的第一属性内容中的最高评分作为第二属性的评分。
  6. 如权利要求1所述的产品推荐方法,其中,所述获取预测产品的第二属性以及第二属性中的第二属性内容,根据第一属性内容的评分以及所述第二属性内容获得用户对预测产品的第二属性的预测评分的步骤包括:
    获取预测产品的第二属性以及第二属性中的第二属性内容,在第二属性中只具有一个第二属性内容时,将对应第二属性内容的第一属性内容的评分作为第二属性的评分,以及;
    获取预测产品的第二属性以及第二属性中的第二属性内容,在第二属性中具有多个第二属性内容时,取对应第二属性内容的第一属性内容的评分的平均值作为第二属性的评分。
  7. 如权利要求6所述的产品推荐方法,其中,所述根据预测产品的评分高低针对所述用户进行产品推荐的步骤包括:
    根据预测产品的评分高低针对所述用户推荐大于预设推荐评分值的预测产品。
  8. 如权利要求6所述的产品推荐方法,其中,所述根据预测产品的评分高低针对所述用户进行产品推荐的步骤包括:
    根据预测产品的评分高低针对所述用户推荐预设推荐个数的预测产品。
  9. 一种产品推荐系统,其中,所述产品推荐系统包括:存储器、处理器及存储在所述存储器上并可在所述处理器上运行的产品推荐程序,所述产品推荐程序被所述处理器执行时实现以下步骤:
    根据多个用户对同类型中多个产品的历史评分,获取多个所述产品的第一属性中第一属性内容的评分;
    获取预测产品的第二属性以及第二属性中的第二属性内容,根据第一属性内容的评分以及所述第二属性内容获得用户对预测产品的第二属性的预测评分;
    将多个用户对同类型中多个产品的历史评分作为协同滤波系统的输入,获取预测产品的协同预测评分;
    将第一属性内容的评分以及所述历史评分作为深度神经网络模型的输入,获取第二属性的预测评分占比以及协同预测评分占比;
    根据第二属性的预测评分、所述第二属性的预测评分占比、协同预测评分以及协同预测评分占比计算多个所述用户对应的所述预测产品的评分,以及;
    根据预测产品的评分高低针对所述用户进行产品推荐。
  10. 如权利要求9所述的产品推荐系统,其中,所述产品推荐程序被所述处理器执行所述根据多个用户对同类型中多个产品的历史评分,获取多个所述产品的第一属性中第一属性内容的评分的步骤包括:
    根据用户对同类型中多个产品的历史评分,在只有一个所述产品中具有第一属性内容时,将具有第一属性内容的该产品的历史评分作为所述第一属性内容的评分,以及;
    根据用户对同类型中多个产品的历史评分,在多个所述产品中均具有第一属性内容时,将具有第一属性内容的多个产品的历史评分求和作为所述第一属性内容的评分。
  11. 如权利要求9所述的产品推荐系统,其中,所述产品推荐程序被所述处理器执行所述根据多个用户对同类型中多个产品的历史评分,获取多个所述产品的第一属性中第一属性内容的评分的步骤包括:
    根据用户对同类型中多个产品的历史评分,在只有一个所述产品中具有第一属性内容时,将具有第一属性内容的该产品的历史评分作为所述第一属性内容的评分,以及;
    根据用户对同类型中多个产品的历史评分,在多个所述产品中均具有第一属性内容时,将具有第一属性内容的多个产品的历史评分取平均值作为所述第一属性内容的评分。
  12. 如权利要求9所述的产品推荐系统,其中,所述产品推荐程序被所述处理器执行所述根据多个用户对同类型中多个产品的历史评分,获取多个所述产品的第一属性中第一属性内容的评分的步骤包括:
    根据用户对同类型中多个产品的历史评分,在只有一个所述产品中具有第一属性内容时,将具有第一属性内容的该产品的历史评分进行归一化处理作为所述第一属性内容的评分,以及;
    根据用户对同类型中多个产品的历史评分,在多个所述产品中均具有第一属性内容时,将具有第一属性内容的多个产品的历史评分取平均值后进行归一化处理作为所述第一属性内容的评分。
  13. 如权利要求9所述的产品推荐系统,其中,所述产品推荐程序被所述处理器执行所述获取预测产品的第二属性以及第二属性中的第二属性内容,根据第一属性内容的评分以及所述第二属性内容获得用户对预测产品的第二属性的预测评分的步骤包括:
    获取预测产品的第二属性以及第二属性中的第二属性内容,在第二属性中只具有一个第二属性内容时,将对应第二属性内容的第一属性内容的评分作为第二属性的评分,以及;
    获取预测产品的第二属性以及第二属性中的第二属性内容,在第二属性中具有多个第二属性内容时,取对应第二属性内容的第一属性内容中的最高评分作为第二属性的评分。
  14. 如权利要求9所述的产品推荐系统,其中,所述产品推荐程序被所述处理器执行所述获取预测产品的第二属性以及第二属性中的第二属性内容,根据第一属性内容的评分以及所述第二属性内容获得用户对预测产品的第二属性的预测评分的步骤包括:
    获取预测产品的第二属性以及第二属性中的第二属性内容,在第二属性中只具有一个第二属性内容时,将对应第二属性内容的第一属性内容的评分作为第二属性的评分,以及;
    获取预测产品的第二属性以及第二属性中的第二属性内容,在第二属性中具有多个第二属性内容时,取对应第二属性内容的第一属性内容的评分的平均值作为第二属性的评分。
  15. 如权利要求14所述的产品推荐系统,其中,所述产品推荐程序被所述处理器执行所述根据预测产品的评分高低针对所述用户进行产品推荐的步骤包括:
    根据预测产品的评分高低针对所述用户推荐大于预设推荐评分值的预测产品。
  16. 如权利要求14所述的产品推荐系统,其中,所述产品推荐程序被所述处理器执行所述根据预测产品的评分高低针对所述用户进行产品推荐的步骤包括:
    根据预测产品的评分高低针对所述用户推荐预设推荐个数的预测产品。
  17. 一种存储介质,其中,所述存储介质上存储有产品推荐程序,所述产品推荐程序被处理器执行时实现以下步骤:
    根据多个用户对同类型中多个产品的历史评分,获取多个所述产品的第一属性中第一属性内容的评分;
    获取预测产品的第二属性以及第二属性中的第二属性内容,根据第一属性内容的评分以及所述第二属性内容获得用户对预测产品的第二属性的预测评分;
    将多个用户对同类型中多个产品的历史评分作为协同滤波系统的输入,获取预测产品的协同预测评分;
    将第一属性内容的评分以及所述历史评分作为深度神经网络模型的输入,获取第二属性的预测评分占比以及协同预测评分占比;
    根据第二属性的预测评分、所述第二属性的预测评分占比、协同预测评分以及协同预测评分占比计算多个所述用户对应的所述预测产品的评分;
    根据预测产品的评分高低针对所述用户进行产品推荐。
  18. 如权利要求17所述的产品推荐系统,其中,所述产品推荐程序被处理器执行所述根据多个用户对同类型中多个产品的历史评分,获取多个所述产品的第一属性中第一属性内容的评分的步骤包括:
    根据用户对同类型中多个产品的历史评分,在只有一个所述产品中具有第一属性内容时,将具有第一属性内容的该产品的历史评分作为所述第一属性内容的评分;
    根据用户对同类型中多个产品的历史评分,在多个所述产品中均具有第一属性内容时,将具有第一属性内容的多个产品的历史评分求和作为所述第一属性内容的评分。
  19. 如权利要求17所述的产品推荐系统,其中,所述产品推荐程序被处理器执行所述根据多个用户对同类型中多个产品的历史评分,获取多个所述产品的第一属性中第一属性内容的评分的步骤包括:
    根据用户对同类型中多个产品的历史评分,在只有一个所述产品中具有第一属性内容时,将具有第一属性内容的该产品的历史评分作为所述第一属性内容的评分;
    根据用户对同类型中多个产品的历史评分,在多个所述产品中均具有第一属性内容时,将具有第一属性内容的多个产品的历史评分取平均值作为所述第一属性内容的评分。
  20. 如权利要求17所述的产品推荐系统,其中,所述产品推荐程序被处理器执行所述根据多个用户对同类型中多个产品的历史评分,获取多个所述产品的第一属性中第一属性内容的评分的步骤包括:
    根据用户对同类型中多个产品的历史评分,在只有一个所述产品中具有第一属性内容时,将具有第一属性内容的该产品的历史评分进行归一化处理作为所述第一属性内容的评分;
    根据用户对同类型中多个产品的历史评分,在多个所述产品中均具有第一属性内容时,将具有第一属性内容的多个产品的历史评分取平均值后进行归一化处理作为所述第一属性内容的评分。
PCT/CN2019/126386 2018-12-29 2019-12-18 产品推荐方法、产品推荐系统及存储介质 WO2020135189A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
EP19904521.2A EP3905179A4 (en) 2018-12-29 2019-12-18 METHOD, SYSTEM AND STORAGE MEDIUM FOR RECOMMENDATION OF PRODUCTS
US17/258,455 US20210295415A1 (en) 2018-12-29 2019-12-18 Product recommendation method, product recommendation system and computer readable storage medium

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201811654674.8 2018-12-29
CN201811654674.8A CN109754316B (zh) 2018-12-29 2018-12-29 产品推荐方法、产品推荐系统及存储介质

Publications (1)

Publication Number Publication Date
WO2020135189A1 true WO2020135189A1 (zh) 2020-07-02

Family

ID=66405130

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/126386 WO2020135189A1 (zh) 2018-12-29 2019-12-18 产品推荐方法、产品推荐系统及存储介质

Country Status (4)

Country Link
US (1) US20210295415A1 (zh)
EP (1) EP3905179A4 (zh)
CN (1) CN109754316B (zh)
WO (1) WO2020135189A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112365159A (zh) * 2020-11-11 2021-02-12 福建亿榕信息技术有限公司 一种基于深度神经网络的后备干部推荐方法及系统

Families Citing this family (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109754316B (zh) * 2018-12-29 2022-07-29 深圳Tcl新技术有限公司 产品推荐方法、产品推荐系统及存储介质
CN111933302B (zh) * 2020-10-09 2021-01-05 平安科技(深圳)有限公司 药物推荐方法、装置、计算机设备及存储介质
US11544315B2 (en) * 2020-10-20 2023-01-03 Spotify Ab Systems and methods for using hierarchical ordered weighted averaging for providing personalized media content
US11693897B2 (en) 2020-10-20 2023-07-04 Spotify Ab Using a hierarchical machine learning algorithm for providing personalized media content
CN113592588B (zh) * 2021-07-25 2023-10-03 深圳市瀚力科技有限公司 基于大数据协同过滤技术的电商平台商品推荐系统及方法
CN113762609A (zh) * 2021-08-27 2021-12-07 浙江天垂科技有限公司 产品质量预测方法及装置

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120143718A1 (en) * 2010-12-03 2012-06-07 Choicestream, Inc. Optimization of a web-based recommendation system
US20140059055A1 (en) * 2012-08-27 2014-02-27 Opera Solutions, Llc System and Method for Combining Multiple Recommender Systems
CN106651542A (zh) * 2016-12-31 2017-05-10 珠海市魅族科技有限公司 一种物品推荐的方法及装置
CN106779825A (zh) * 2016-12-02 2017-05-31 乐视控股(北京)有限公司 一种物品推荐方法、装置及电子设备
CN108874914A (zh) * 2018-05-29 2018-11-23 吉林大学 一种基于图卷积与神经协同过滤的信息推荐方法
CN108959603A (zh) * 2018-07-13 2018-12-07 北京印刷学院 基于深度神经网络的个性化推荐系统及方法
CN109754316A (zh) * 2018-12-29 2019-05-14 深圳Tcl新技术有限公司 产品推荐方法、产品推荐系统及存储介质

Family Cites Families (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5867799A (en) * 1996-04-04 1999-02-02 Lang; Andrew K. Information system and method for filtering a massive flow of information entities to meet user information classification needs
US7937725B1 (en) * 2000-07-27 2011-05-03 Koninklijke Philips Electronics N.V. Three-way media recommendation method and system
US8966394B2 (en) * 2008-09-08 2015-02-24 Apple Inc. System and method for playlist generation based on similarity data
KR20140140309A (ko) * 2013-05-29 2014-12-09 (주) 다이퀘스트 사용자 간의 유사도 계산방법 및 이를 이용한 아이템 추천방법
CN103559622A (zh) * 2013-07-31 2014-02-05 焦点科技股份有限公司 基于特征的协同过滤推荐方法
US9978362B2 (en) * 2014-09-02 2018-05-22 Microsoft Technology Licensing, Llc Facet recommendations from sentiment-bearing content
CN107491540A (zh) * 2017-08-24 2017-12-19 济南浚达信息技术有限公司 一种结合深度贝叶斯模型和协同异构信息嵌入的电影推荐方法
CN108334592B (zh) * 2018-01-30 2021-11-02 南京邮电大学 一种基于内容与协同过滤相结合的个性化推荐方法
US11551280B2 (en) * 2018-11-01 2023-01-10 Netflix, Inc. Method, manufacture, and system for recommending items to users

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120143718A1 (en) * 2010-12-03 2012-06-07 Choicestream, Inc. Optimization of a web-based recommendation system
US20140059055A1 (en) * 2012-08-27 2014-02-27 Opera Solutions, Llc System and Method for Combining Multiple Recommender Systems
CN106779825A (zh) * 2016-12-02 2017-05-31 乐视控股(北京)有限公司 一种物品推荐方法、装置及电子设备
CN106651542A (zh) * 2016-12-31 2017-05-10 珠海市魅族科技有限公司 一种物品推荐的方法及装置
CN108874914A (zh) * 2018-05-29 2018-11-23 吉林大学 一种基于图卷积与神经协同过滤的信息推荐方法
CN108959603A (zh) * 2018-07-13 2018-12-07 北京印刷学院 基于深度神经网络的个性化推荐系统及方法
CN109754316A (zh) * 2018-12-29 2019-05-14 深圳Tcl新技术有限公司 产品推荐方法、产品推荐系统及存储介质

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP3905179A4 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112365159A (zh) * 2020-11-11 2021-02-12 福建亿榕信息技术有限公司 一种基于深度神经网络的后备干部推荐方法及系统

Also Published As

Publication number Publication date
EP3905179A1 (en) 2021-11-03
CN109754316B (zh) 2022-07-29
US20210295415A1 (en) 2021-09-23
EP3905179A4 (en) 2022-09-14
CN109754316A (zh) 2019-05-14

Similar Documents

Publication Publication Date Title
WO2020135189A1 (zh) 产品推荐方法、产品推荐系统及存储介质
WO2021052085A1 (zh) 视频推荐方法、装置、电子设备及计算机可读介质
JP7141410B2 (ja) 空間的3d環境に対するコンテンツのマッチング
US11127074B2 (en) Recommendations based on object detected in an image
US10346684B2 (en) Visual search utilizing color descriptors
US20210084360A1 (en) Methods and systems for displaying content
US20160378762A1 (en) Methods and systems for identifying media assets
AU2018204372A1 (en) A united framework for personalisation which encompasses cold start and implicit interaction data
US20170127102A1 (en) Method and electronic device for video recommendation
US9411839B2 (en) Index configuration for searchable data in network
US10630792B2 (en) Methods and systems for viewing user feedback
US11055759B1 (en) Color selection for image matching visual search
US10311500B2 (en) Methods and systems for developer onboarding for software-development products
WO2016201800A1 (zh) 信息获取方法、服务器、终端及数据库构建方法、装置
US20170235828A1 (en) Text Digest Generation For Searching Multiple Video Streams
CN112364203B (zh) 电视视频推荐方法、装置、服务器和存储介质
US9111231B2 (en) Associating a web session with a household member
US20150189384A1 (en) Presenting information based on a video
CN109121000A (zh) 一种视频处理方法及客户端
WO2019071890A1 (zh) 产品推荐装置、方法及计算机可读存储介质
CN104346431A (zh) 信息处理装置、信息处理方法和程序
WO2020037886A1 (zh) 视频关联方法、装置及计算机可读存储介质
WO2015090133A1 (zh) 一种视频信息更新方法及电子设备
US20140289334A1 (en) System and method for recommending multimedia information
US8984091B1 (en) Providing content based on timestamp of last request for content

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19904521

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

WWE Wipo information: entry into national phase

Ref document number: 2019904521

Country of ref document: EP

ENP Entry into the national phase

Ref document number: 2019904521

Country of ref document: EP

Effective date: 20210729