WO2016101302A1 - User log and entity relationship gallery-based personalized recommendation system and recommendation method thereof - Google Patents
User log and entity relationship gallery-based personalized recommendation system and recommendation method thereof Download PDFInfo
- Publication number
- WO2016101302A1 WO2016101302A1 PCT/CN2014/095495 CN2014095495W WO2016101302A1 WO 2016101302 A1 WO2016101302 A1 WO 2016101302A1 CN 2014095495 W CN2014095495 W CN 2014095495W WO 2016101302 A1 WO2016101302 A1 WO 2016101302A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- user
- data
- entity
- recommendation
- attributes
- Prior art date
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
Definitions
- the invention relates to the technical field of recommendation systems based on big data and data mining, in particular to a personalized recommendation system based on user logs and entity association galleries and a recommendation method thereof.
- the most widely used recommendation system is collaborative filtering technology, including neighbor-based and model-based methods.
- the neighbor-based method is usually used to study the similarity calculation problem between users or items: the user-based collaborative filtering method first finds users with similar main attributes (ie, neighbors), and then collects and analyzes the neighbor's data to further recommend to the target users.
- the items of interest Similarly, item-based collaborative filtering methods take advantage of the information rating of similar items.
- the model-based approach attempts to transform user-item relationship assessment data into different models (such as Bayesian networks, factorization or cluster models, etc.) and use these models to recommend items to users in unknown scenarios (decomposition technique) Get more and more attention in applications with explicit or implicit feedback).
- the invention provides a personalized recommendation system based on a user log and an entity association library; the cold start problem of the recommendation system can be effectively solved, and the recommendation result dynamically changes with the conversion of the user preference, and the user experience of the recommendation system is enhanced. Sexual and individual needs.
- Another technical problem to be solved by the present invention is to provide a recommendation method of a personalized recommendation system based on a user log and an entity association library, effectively solving the cold start problem of the recommendation system, and dynamically changing the recommendation result according to the conversion of the user preference. Enhance user experience and personalization requirements for the recommendation system.
- the system mainly includes the following modules:
- the data interface layer is mainly responsible for data communication with the service layer, receiving data requests, and transmitting recommendation results;
- a knowledge base as a data source for the system and a learning set for the recommendation computing system
- User log system which records all operation records of the user at the business layer
- the entity association library analyzes the data of the knowledge base, analyzes and stores the relationship between the user, the entity data, and the attributes;
- the computing system is recommended to receive the input of the data interface layer, calculate the recent preference value according to the user log system, and integrate the preference data of the nearest neighbor to obtain a sequence of data entities that the user may be most interested in;
- the data entities of the knowledge base are mainly textual contents such as news and literature, and the amount of data continues to grow over time.
- the entity relationship library is by analyzing each entity data of the knowledge base and user configuration File, decompose its attributes, calculate its own weights (with a range of [0, 1]) through TF-IDF and other related algorithms, and record entity-attribute, attribute-attribute, entity-entity, user-user, in multidimensional graph. User-entity, user-attribute, etc. are related to each other.
- the user log system records all the operations of the user of the service layer, but when applied to the recommended computing system, the user mainly extracts the click operation log of the user to the entity data, and calculates the user according to the association relationship of the entity data between different time segments.
- a relatively stable preference value at some point in time as an input to a personalized recommendation.
- Mainly includes the following steps,
- Step 1 When the user clicks to view an entity data, the associated weight of the corresponding attribute in the entity association gallery will also dynamically change and be recalculated;
- Step 2 analyzing the correlation between the time point of the user currently clicking the entity data and the entity data between the earlier time points, and continuously finding the nearest time point with the highest correlation value by successive iteration convergence, taking the two time points Inter-entity data as the user's most recent preference value data source;
- Step 3 According to the user's recent preference value and the dynamically changing entity association library, the comprehensive processing results in the recommendation result closest to the user's current preference.
- Figure 1 is a frame structure of the personalized recommendation system of the present invention
- FIG. 2 is a schematic diagram of an entity relationship library of the personalized recommendation system of the present invention.
- Figure 3 is a flow chart of the preferred method of the present invention.
- a personalized recommendation system based on a user log and an entity association library mainly includes but is not limited to five parts: a data interface layer, a knowledge base, a user log system, an entity relationship library, and a recommendation operation. system.
- the data interface layer is mainly responsible for data communication with the service layer, receiving data requests and sending entity data recommendation results;
- the knowledge base is a learning source of the application system data source and the recommendation computing system;
- the user log system records the user at the business layer. All operation records; entity relationship library, separating the data entity attribute and user attribute of the knowledge base and calculating and storing the relationship between them; recommending the computing system, receiving the input of the data interface layer, calculating the preference according to the user log system,
- the entity relational database finds or analyzes the relationship of its entity attributes to obtain a sequence of data entities with the strongest correlation.
- the entity relationship library diagram of the personalized recommendation system based on the user log and the entity association library is to analyze the correlation of the data in the knowledge base from the horizontal and vertical angles.
- Each attribute of the entity data and the user has its own global weight ⁇ (mainly through the correlation algorithm such as TF-IDF to calculate its global word frequency), and the different attributes of the entity data and the user have a proportion according to the importance degree of the attribute.
- the parameter ⁇ whereby the entity data or the weight ⁇ between the users can be separately calculated; and the user's recent preference value parameter ⁇ is calculated according to the user log record, and the nearest neighbor user can be found based on the ⁇ parameter to obtain the preference data. As a result of the recommendation.
- a flow chart of the steps of the personalized method based on the user log and the entity association library can effectively solve the problem of cold start and user preference drift of the recommendation system.
- the method includes the following steps:
- the user views the detailed information of an entity data, and at this time, it is necessary to determine whether the user is a new user;
- the nearest neighbor user of the user is searched through the entity association graph, and the preference data of the nearest neighbor user is used as the recommendation result;
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Databases & Information Systems (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Information Transfer Between Computers (AREA)
Abstract
The present invention relates to the technical field of big data and data mining recommendation systems, and in particular to a user log and entity relationship gallery-based personalized recommendation system and a recommendation method thereof. The system of the present invention comprises: a data interface layer, a user log system, a knowledge base, an entity relationship gallery and a recommendation operational system. The data interface layer is used for communicating with upper layer business systems; the user log system comprises all operation records of a user in an application system; the knowledge base is a set of all data in the application system and a learning set of the recommendation system; the entity relationship gallery is used for storing correlation relationships among the user, data entities, attributes and the like; and the recommendation operational system automatically recommends topic data, in which the user is interested, to the user according to a particular algorithm by integrating user preference and user weight. The present invention solves the problem of cold start of the recommendation system and also the problem that the recommendation operational complexity increases because of the constantly changing interest of the user; the personalized recommendation system and method can be applied to big data processing.
Description
本发明涉及基于大数据和数据挖掘推荐系统技术领域，尤其是一种基于用户日志和实体关联图库的个性化推荐系统及其推荐方法。The invention relates to the technical field of recommendation systems based on big data and data mining, in particular to a personalized recommendation system based on user logs and entity association galleries and a recommendation method thereof.
近年来，随着信息技术的日益发展以及连入互联网用户数量的不断增长，互联网中时时刻刻都会产生数量巨大的比特数据。如何有效利用这些信息已经成为一个世界性关注的问题。在经过学术界和工业界多年的研究和应用，对于这些数据附加价值的挖掘和利用技术也逐渐成熟，最主要的就是在电子商务、在线信息应用(如在线新闻、在线音乐、在线影视)等领域的关联信息推荐系统。In recent years, with the development of information technology and the continuous increase in the number of users connected to the Internet, the Internet has generated a huge amount of bit data at all times. How to effectively use this information has become a worldwide concern. After years of research and application in academia and industry, the technology of excavating and utilizing the added value of these data has gradually matured. The most important ones are in e-commerce and online information applications (such as online news, online music, online video). A related information recommendation system for the field.
推荐系统里应用最广泛的是协同过滤技术，包括基于邻居和基于模型的方法。基于邻居的方法通常用于研究用户或物品之间的相似度计算问题：基于用户的协同过滤方法首先查找主要属性相似的用户(即邻居)，然后收集和分析邻居的数据以进一步为目标用户推荐其感兴趣的物品。相似的，基于物品的协同过滤方法利用了相似物品的信息评级的优势。基于模型的方法试图将用户-物品的关系评估数据转化为不同的模型(如贝叶斯网络、因式分解或者集群模型等)并使用这些模型在未知场景中向用户推荐物品(因式分解技术在具有显式或隐式反馈的应用系统中获得了越来越高的关注)。The most widely used recommendation system is collaborative filtering technology, including neighbor-based and model-based methods. The neighbor-based method is usually used to study the similarity calculation problem between users or items: the user-based collaborative filtering method first finds users with similar main attributes (ie, neighbors), and then collects and analyzes the neighbor's data to further recommend to the target users. The items of interest. Similarly, item-based collaborative filtering methods take advantage of the information rating of similar items. The model-based approach attempts to transform user-item relationship assessment data into different models (such as Bayesian networks, factorization or cluster models, etc.) and use these models to recommend items to users in unknown scenarios (decomposition technique) Get more and more attention in applications with explicit or implicit feedback).
然而目前主流的推荐系统在面对新用户时，由于没有历史数据作为推荐依据而带来的“冷启动”问题也让用户体验比较差，而且用户的数据阅读兴趣或偏好也是随时问而不断变化，如何比较好的解决推荐系统的冷启动问题并且随
时根据用户的偏好变化而调整推荐结果也是一个研究和应用的热点话题。However, the current mainstream recommendation system in the face of new users, because of the lack of historical data as a basis for recommendation, the "cold start" problem also makes the user experience worse, and the user's data reading interest or preference is constantly changing and changing. How to better solve the cold start problem of the recommendation system and
Adjusting the recommendation results according to changes in user preferences is also a hot topic of research and application.
发明内容Summary of the invention
本发明提供了一种基于用户日志和实体关联图库的个性化推荐系统；可以有效解决推荐系统的冷启动问题，并且使推荐结果随用户偏好的转换而动态变化，增强用户对推荐系统的使用体验性和个性化需求。The invention provides a personalized recommendation system based on a user log and an entity association library; the cold start problem of the recommendation system can be effectively solved, and the recommendation result dynamically changes with the conversion of the user preference, and the user experience of the recommendation system is enhanced. Sexual and individual needs.
本发明解决的另一技术问题在于提供一种基于用户日志和实体关联图库的个性化推荐系统的推荐方法，有效解决推荐系统的冷启动问题，并且使推荐结果随用户偏好的转换而动态变化，增强用户对推荐系统的使用体验性和个性化需求。Another technical problem to be solved by the present invention is to provide a recommendation method of a personalized recommendation system based on a user log and an entity association library, effectively solving the cold start problem of the recommendation system, and dynamically changing the recommendation result according to the conversion of the user preference. Enhance user experience and personalization requirements for the recommendation system.
本发明解决上述技术问题之一的技术方案是：The technical solution of the present invention to solve one of the above technical problems is:
所述的系统主要包括有以下模块：The system mainly includes the following modules:
数据接口层，主要负责与业务层的数据通信，接收数据请求以及发送推荐结果；The data interface layer is mainly responsible for data communication with the service layer, receiving data requests, and transmitting recommendation results;
知识库，作为系统的数据源和推荐运算系统的学习集；a knowledge base, as a data source for the system and a learning set for the recommendation computing system;
用户日志系统，记录用户在业务层的所有操作记录；User log system, which records all operation records of the user at the business layer;
实体关联图库，分析知识库的数据，剖析并存储用户、实体数据、属性相互之间的关联关系；The entity association library analyzes the data of the knowledge base, analyzes and stores the relationship between the user, the entity data, and the attributes;
推荐运算系统，接收数据接口层的输入，根据用户日志系统计算其最近偏好值，综合最邻近用户的偏好数据，得出用户可能最感兴趣的一个数据实体序列；The computing system is recommended to receive the input of the data interface layer, calculate the recent preference value according to the user log system, and integrate the preference data of the nearest neighbor to obtain a sequence of data entities that the user may be most interested in;
所述的知识库的数据实体主要为新闻、文献等文字性内容，且数据量随时间推移而不断增长。The data entities of the knowledge base are mainly textual contents such as news and literature, and the amount of data continues to grow over time.
所述实体关系图库是通过分析所述知识库的每一项实体数据以及用户配置
文件，分解其属性，通过TF-IDF等相关算法计算各自的权值(范围为[0，1])，并以多维图形式记录实体-属性、属性-属性、实体-实体、用户-用户、用户-实体、用户-属性等相互之间的关联关系。The entity relationship library is by analyzing each entity data of the knowledge base and user configuration
File, decompose its attributes, calculate its own weights (with a range of [0, 1]) through TF-IDF and other related algorithms, and record entity-attribute, attribute-attribute, entity-entity, user-user, in multidimensional graph. User-entity, user-attribute, etc. are related to each other.
所述用户日志系统记录了业务层用户所有的操作，但在应用到推荐运算系统时，主要是提取用户对实体数据的点击操作日志，并且根据不同时间段之间实体数据的关联关系计算用户在某个时间点相对稳定的偏好值以作为个性化推荐的输入。The user log system records all the operations of the user of the service layer, but when applied to the recommended computing system, the user mainly extracts the click operation log of the user to the entity data, and calculates the user according to the association relationship of the entity data between different time segments. A relatively stable preference value at some point in time as an input to a personalized recommendation.
本发明解决上述技术问题之二的技术方案是：The technical solution of the present invention to solve the above two technical problems is:
主要包括如下步骤，Mainly includes the following steps,
步骤一，用户点击查看一条实体数据时，实体关联图库中相应属性的关联权值也会动态变化并重新被计算；Step 1: When the user clicks to view an entity data, the associated weight of the corresponding attribute in the entity association gallery will also dynamically change and be recalculated;
步骤二，分析用户当前点击实体数据的时间点和一个较早的时间点之间的实体数据的相关性，通过不断迭代收敛找出相关值最高的一个最近时间点，取这两个时间点之间的实体数据作为用户的最近偏好值数据源；Step 2: analyzing the correlation between the time point of the user currently clicking the entity data and the entity data between the earlier time points, and continuously finding the nearest time point with the highest correlation value by successive iteration convergence, taking the two time points Inter-entity data as the user's most recent preference value data source;
步骤三，根据用户的最近偏好值和动态变化的实体关联图库，综合处理后得出与用户当前偏好最接近的推荐结果。Step 3: According to the user's recent preference value and the dynamically changing entity association library, the comprehensive processing results in the recommendation result closest to the user's current preference.
所述的步骤二中计算两个指定时间点范围内实体数据的相关性时采取基于局部敏感哈希的随机投影方法，时间复杂度从传统的K-NN方法的O(N2)变为O(N)。When the correlation of the entity data in the two specified time points is calculated in the second step, a local-sensitive hash-based random projection method is adopted, and the time complexity is changed from O(N 2 ) of the traditional K-NN method to O. (N).
下面结合附图对本发明进一步说明：The present invention is further described below in conjunction with the accompanying drawings:
图1是本发明个性化推荐系统的框架结构；
Figure 1 is a frame structure of the personalized recommendation system of the present invention;
图2是本发明个性化推荐系统的实体关系图库示意图；2 is a schematic diagram of an entity relationship library of the personalized recommendation system of the present invention;
图3是本发明推荐方法的流程图。Figure 3 is a flow chart of the preferred method of the present invention.
如附图1所示，本发明一种基于基于用户日志和实体关联图库的个性化推荐系统主要包括但不限于5个部分：数据接口层、知识库、用户日志系统、实体关系图库、推荐运算系统。As shown in FIG. 1 , a personalized recommendation system based on a user log and an entity association library mainly includes but is not limited to five parts: a data interface layer, a knowledge base, a user log system, an entity relationship library, and a recommendation operation. system.
其中数据接口层主要负责与业务层的数据通信，接收数据请求以及发送实体数据推荐结果；知识库，是应用系统的数据源和推荐运算系统的学习集；用户日志系统，记录用户在业务层的所有操作记录；实体关系图库，将知识库的数据实体属性和用户属性分离并计算、存储相互之间的关联关系；推荐运算系统，接收数据接口层的输入，根据用户日志系统计算其偏好，在实体关系数据库查找或分析其实体属性关系，得出相关性最强的一个数据实体序列。The data interface layer is mainly responsible for data communication with the service layer, receiving data requests and sending entity data recommendation results; the knowledge base is a learning source of the application system data source and the recommendation computing system; the user log system records the user at the business layer. All operation records; entity relationship library, separating the data entity attribute and user attribute of the knowledge base and calculating and storing the relationship between them; recommending the computing system, receiving the input of the data interface layer, calculating the preference according to the user log system, The entity relational database finds or analyzes the relationship of its entity attributes to obtain a sequence of data entities with the strongest correlation.
如附图2所示，本发明一种基于用户日志和实体关联图库的个性化推荐系统的实体关系图库示意图是同时从横向和纵向的角度剖析知识库中数据的相互关联性。As shown in FIG. 2, the entity relationship library diagram of the personalized recommendation system based on the user log and the entity association library is to analyze the correlation of the data in the knowledge base from the horizontal and vertical angles.
实体数据和用户的每个属性都有其在全局的一个权值ω(主要是通过TF-IDF等相关算法计算其全局词频)，同时实体数据和用户的不同属性又根据属性重要程度有一个比重参数θ，由此可分别计算出实体数据或者用户之间的权值ω；同时根据用户日志记录计算出用户的最近偏好值参数λ，可以以λ参数为依据寻找最邻近用户，获取其偏好数据作为推荐结果。Each attribute of the entity data and the user has its own global weight ω (mainly through the correlation algorithm such as TF-IDF to calculate its global word frequency), and the different attributes of the entity data and the user have a proportion according to the importance degree of the attribute. The parameter θ, whereby the entity data or the weight ω between the users can be separately calculated; and the user's recent preference value parameter λ is calculated according to the user log record, and the nearest neighbor user can be found based on the λ parameter to obtain the preference data. As a result of the recommendation.
如附图3所示，本发明一种基于用户日志和实体关联图库的个性化方法的步骤流程图，可以有效解决推荐系统的冷启动和用户偏好漂移问题。该方法包括以下步骤：
As shown in FIG. 3, a flow chart of the steps of the personalized method based on the user log and the entity association library can effectively solve the problem of cold start and user preference drift of the recommendation system. The method includes the following steps:
1、用户查看某个实体数据的详细信息，此时需判断用户是否为新用户；1. The user views the detailed information of an entity data, and at this time, it is necessary to determine whether the user is a new user;
2、若用户为新用户，通过实体关联图查找该用户的最邻近用户，以最邻近用户的偏好数据作为推荐结果；2. If the user is a new user, the nearest neighbor user of the user is searched through the entity association graph, and the preference data of the nearest neighbor user is used as the recommendation result;
3、若用户为老用户，首先在日志系统中以当前的日志记录为终点分析并计算用户的最近稳定偏好区间数据，得出偏好值，然后综合该偏好值以及最邻近用户权值得出关联度最高的实体数据作为推荐结果。
3. If the user is an old user, first analyze and calculate the user's recent stable preference interval data in the log system with the current log record as the end point, and obtain the preference value, and then integrate the preference value and the nearest neighbor user weight to be associated. The highest entity data is used as the recommendation result.
Claims (9)
- 一种基于用户日志和实体关联图库的个性化推荐系统，其特征在于：所述的系统主要包括有以下模块：A personalized recommendation system based on a user log and an entity association library, characterized in that the system mainly comprises the following modules:数据接口层，主要负责与业务层的数据通信，接收数据请求以及发送推荐结果；The data interface layer is mainly responsible for data communication with the service layer, receiving data requests, and transmitting recommendation results;知识库，作为系统的数据源和推荐运算系统的学习集；a knowledge base, as a data source for the system and a learning set for the recommendation computing system;用户日志系统，记录用户在业务层的所有操作记录；User log system, which records all operation records of the user at the business layer;实体关联图库，分析知识库的数据，剖析并存储用户、实体数据、属性相互之间的关联关系；The entity association library analyzes the data of the knowledge base, analyzes and stores the relationship between the user, the entity data, and the attributes;推荐运算系统，接收数据接口层的输入，根据用户日志系统计算其最近偏好值，综合最邻近用户的偏好数据，得出用户可能最感兴趣的一个数据实体序列；The computing system is recommended to receive the input of the data interface layer, calculate the recent preference value according to the user log system, and integrate the preference data of the nearest neighbor to obtain a sequence of data entities that the user may be most interested in;
- 根据权利要求1所述的个性化推荐系统，其特征在于：所述的知识库的数据实体主要为新闻、文献等文字性内容，且数据量随时间推移而不断增长。The personalized recommendation system according to claim 1, wherein the data entity of the knowledge base is mainly textual content such as news and documents, and the amount of data is continuously increased over time.
- 根据权利要求1所述的个性化推荐系统，其特征在于：所述实体关系图库是通过分析所述知识库的每一项实体数据以及用户配置文件，分解其属性，通过TF-IDF等相关算法计算各自的权值(范围为[0，1])，并以多维图形式记录实体-属性、属性-属性、实体-实体、用户-用户、用户-实体、用户-属性等相互之间的关联关系。The personalized recommendation system according to claim 1, wherein the entity relationship library is configured by analyzing each entity data and user profile of the knowledge base, decomposing attributes thereof, and using a related algorithm such as TF-IDF. Calculate the respective weights (range [0,1]) and record the associations between entity-attributes, attribute-attributes, entity-entities, user-users, user-entities, user-attributes, etc. in a multi-dimensional graph relationship.
- 根据权利要求2所述的个性化推荐系统，其特征在于：所述实体关系图库是通过分析所述知识库的每一项实体数据以及用户配置文件，分解其属性， 通过TF-IDF等相关算法计算各自的权值(范围为[0，1])，并以多维图形式记录实体-属性、属性-属性、实体-实体、用户-用户、用户-实体、用户-属性等相互之间的关联关系。The personalized recommendation system according to claim 2, wherein the entity relationship library is decomposed by analyzing each entity data of the knowledge base and a user profile. Calculate their respective weights (range [0,1]) by TF-IDF and other related algorithms, and record entity-attribute, attribute-attribute, entity-entity, user-user, user-entity, user-in multidimensional graph The relationship between attributes and the like.
- 根据权利要求1至4任一项所述的个性化推荐系统，其特征在于：所述用户日志系统记录了业务层用户所有的操作，但在应用到推荐运算系统时，主要是提取用户对实体数据的点击操作日志，并且根据不同时间段之间实体数据的关联关系计算用户在某个时间点相对稳定的偏好值以作为个性化推荐的输入。The personalized recommendation system according to any one of claims 1 to 4, wherein the user log system records all operations of the business layer user, but when applied to the recommendation computing system, the user is mainly extracted from the entity. The click operation log of the data, and the relative value of the user's relative stability at a certain point in time is calculated as an input of the personalized recommendation according to the association relationship of the entity data between different time periods.
- 根据权利要求1至4任一项所述的个性化推荐系统，其特征在于：所述推荐运算系统中，用户第一次在业务层获取推荐数据时是用户本身的配置文件，此时主要采取基于用户的协同过滤方法获取该用户的最邻近用户之最热实体数据为推荐结果；正常情况下还需综合考虑用户以往的历史点击记录(相对偏好值)，减少实体数据的稀疏性问题对结果的影响。The personalized recommendation system according to any one of claims 1 to 4, wherein in the recommendation computing system, when the user first obtains the recommended data in the service layer, the user is the configuration file of the user, and the main adoption is The user-based collaborative filtering method obtains the hottest entity data of the nearest neighbor user of the user as a recommendation result; under normal circumstances, the user needs to comprehensively consider the historical history click record (relative preference value) of the user, and reduce the sparsity problem of the entity data to the result. Impact.
- 根据权利要求5所述的个性化推荐系统，其特征在于：所述推荐运算系统中，用户第一次在业务层获取推荐数据时是用户本身的配置文件，此时主要采取基于用户的协同过滤方法获取该用户的最邻近用户之最热实体数据为推荐结果；正常情况下还需综合考虑用户以往的历史点击记录(相对偏好值)，减少实体数据的稀疏性问题对结果的影响。The personalized recommendation system according to claim 5, wherein in the recommendation computing system, when the user first obtains the recommendation data in the service layer, the user is a configuration file of the user, and the user-based collaborative filtering is adopted at this time. The method obtains the hottest entity data of the nearest user of the user as a recommendation result; under normal circumstances, the user historical history record (relative preference value) needs to be comprehensively considered to reduce the influence of the sparseness problem of the entity data on the result.
- 一种权利要求1所述的个性化推荐系统的推荐方法，其特征在于：主要包括如下步骤，A method for recommending a personalized recommendation system according to claim 1, characterized in that it mainly comprises the following steps:步骤一，用户点击查看一条实体数据时，实体关联图库中相应属性的关联权值也会动态变化并重新被计算； Step 1: When the user clicks to view an entity data, the associated weight of the corresponding attribute in the entity association gallery will also dynamically change and be recalculated;步骤二，分析用户当前点击实体数据的时间点和一个较早的时间点之间的实体数据的相关性，通过不断迭代收敛找出相关值最高的一个最近时间点，取这两个时间点之间的实体数据作为用户的最近偏好值数据源；Step 2: analyzing the correlation between the time point of the user currently clicking the entity data and the entity data between the earlier time points, and continuously finding the nearest time point with the highest correlation value by successive iteration convergence, taking the two time points Inter-entity data as the user's most recent preference value data source;步骤三，根据用户的最近偏好值和动态变化的实体关联图库，综合处理后得出与用户当前偏好最接近的推荐结果。Step 3: According to the user's recent preference value and the dynamically changing entity association library, the comprehensive processing results in the recommendation result closest to the user's current preference.
- 根据权利要求8所述的推荐方法，其特征在于：所述的步骤二中计算两个指定时间点范围内实体数据的相关性时采取基于局部敏感哈希的随机投影方法，时间复杂度从传统的K-NN方法的O(N2)变为O(N)。 The recommendation method according to claim 8, wherein in the step 2, a local projection-based random projection method is adopted when calculating the correlation of the entity data in the two specified time points, and the time complexity is from the conventional The O(N 2 ) of the K-NN method becomes O(N).
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410831549.5 | 2014-12-25 | ||
| CN201410831549.5A CN104462560B (en) | 2014-12-25 | 2014-12-25 | A kind of recommendation method of personalized recommendation system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| WO2016101302A1 true WO2016101302A1 (en) | 2016-06-30 |
Family
ID=52908595
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| PCT/CN2014/095495 WO2016101302A1 (en) | 2014-12-25 | 2014-12-30 | User log and entity relationship gallery-based personalized recommendation system and recommendation method thereof |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN104462560B (en) |
| WO (1) | WO2016101302A1 (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111104603A (en) * | 2019-11-01 | 2020-05-05 | 复旦大学 | Real-time hybrid recommendation method and system based on Lambda architecture |
| CN111625837A (en) * | 2020-05-22 | 2020-09-04 | 北京金山云网络技术有限公司 | Method and device for identifying system vulnerability and server |
| CN112069408A (en) * | 2020-06-15 | 2020-12-11 | 北京理工大学 | Recommendation system and method for fusion relation extraction |
| CN113283908A (en) * | 2021-06-09 | 2021-08-20 | 武汉斗鱼鱼乐网络科技有限公司 | Target group identification method and device |
Families Citing this family (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104850645B (en) * | 2015-05-28 | 2018-08-14 | 苏州大学张家港工业技术研究院 | A kind of Active Learning scoring bootstrap technique and system based on matrix decomposition |
| CN105426392B (en) * | 2015-10-28 | 2019-03-26 | 浪潮软件集团有限公司 | Collaborative filtering recommendation method and system |
| CN105488214A (en) * | 2015-12-15 | 2016-04-13 | 成都陌云科技有限公司 | Page customization method based on big data |
| CN106251195A (en) * | 2016-07-25 | 2016-12-21 | 四川易想电子商务有限公司 | E-commerce system based on B/S structure |
| CN106097079A (en) * | 2016-07-25 | 2016-11-09 | 四川易想电子商务有限公司 | E-commerce system based on UML |
| CN106204241A (en) * | 2016-07-25 | 2016-12-07 | 四川易想电子商务有限公司 | A kind of electronic commerce transaction system |
| CN106228406A (en) * | 2016-07-25 | 2016-12-14 | 四川易想电子商务有限公司 | Technologies of Recommendation System in E-Commerce based on personalized recommendation |
| CN106503267A (en) * | 2016-12-07 | 2017-03-15 | 电子科技大学 | A kind of personalized recommendation algorithm suitable for user preference dynamic evolution |
| CN107122411B (en) * | 2017-03-29 | 2020-08-14 | 浙江大学 | Collaborative filtering recommendation method based on discrete multi-view Hash |
| CN107066582B (en) * | 2017-04-14 | 2020-06-26 | 聚好看科技股份有限公司 | Method and device for realizing virtual resource recommendation |
| CN112437940A (en) * | 2018-06-01 | 2021-03-02 | 系统软件企业有限责任公司 | System and method for recommendation system based on implicit feedback |
| CN108921657B (en) * | 2018-06-25 | 2021-06-29 | 中国人民大学 | Knowledge-enhanced memory network-based sequence recommendation method |
| CN108965844A (en) * | 2018-07-06 | 2018-12-07 | 维沃移动通信有限公司 | One kind is taken pictures effect adjusting method, device and mobile terminal |
| CN110083773B (en) * | 2019-04-30 | 2022-09-23 | 南京邮电大学 | Personalized knowledge recommendation method and computer storage medium |
| CN111461118B (en) * | 2020-03-31 | 2023-11-24 | 中国移动通信集团黑龙江有限公司 | Interest feature determining method, device, equipment and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1606746A (en) * | 2000-11-22 | 2005-04-13 | 皇家菲利浦电子有限公司 | Television program recommender with interval-based profiles for determining time-varying conditional probabilities |
| CN102402766A (en) * | 2011-12-27 | 2012-04-04 | 纽海信息技术（上海）有限公司 | User interest modeling method based on web page browsing |
| US20130013622A1 (en) * | 2011-07-09 | 2013-01-10 | Yahoo!, Inc. | Media Recommendations for a Social-Software Website |
| CN103049865A (en) * | 2012-12-17 | 2013-04-17 | 中国农业大学 | Method and system for initiatively recommending product information service |
| CN103544663A (en) * | 2013-06-28 | 2014-01-29 | Tcl集团股份有限公司 | Method and system for recommending network public classes and mobile terminal |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102208086B (en) * | 2010-03-31 | 2014-05-14 | 北京邮电大学 | Field-oriented personalized intelligent recommendation system and implementation method |
| KR20130035064A (en) * | 2011-09-29 | 2013-04-08 | 삼성전자주식회사 | Method and system for providing personalization service based on personal tendency |
| CN103823908B (en) * | 2014-03-21 | 2017-07-14 | 北京飞流九天科技有限公司 | Content recommendation method and server based on user preference |
-
2014
- 2014-12-25 CN CN201410831549.5A patent/CN104462560B/en active Active
- 2014-12-30 WO PCT/CN2014/095495 patent/WO2016101302A1/en active Application Filing
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1606746A (en) * | 2000-11-22 | 2005-04-13 | 皇家菲利浦电子有限公司 | Television program recommender with interval-based profiles for determining time-varying conditional probabilities |
| US20130013622A1 (en) * | 2011-07-09 | 2013-01-10 | Yahoo!, Inc. | Media Recommendations for a Social-Software Website |
| CN102402766A (en) * | 2011-12-27 | 2012-04-04 | 纽海信息技术（上海）有限公司 | User interest modeling method based on web page browsing |
| CN103049865A (en) * | 2012-12-17 | 2013-04-17 | 中国农业大学 | Method and system for initiatively recommending product information service |
| CN103544663A (en) * | 2013-06-28 | 2014-01-29 | Tcl集团股份有限公司 | Method and system for recommending network public classes and mobile terminal |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111104603A (en) * | 2019-11-01 | 2020-05-05 | 复旦大学 | Real-time hybrid recommendation method and system based on Lambda architecture |
| CN111625837A (en) * | 2020-05-22 | 2020-09-04 | 北京金山云网络技术有限公司 | Method and device for identifying system vulnerability and server |
| CN112069408A (en) * | 2020-06-15 | 2020-12-11 | 北京理工大学 | Recommendation system and method for fusion relation extraction |
| CN113283908A (en) * | 2021-06-09 | 2021-08-20 | 武汉斗鱼鱼乐网络科技有限公司 | Target group identification method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104462560B (en) | 2018-01-05 |
| CN104462560A (en) | 2015-03-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2016101302A1 (en) | User log and entity relationship gallery-based personalized recommendation system and recommendation method thereof | |
| CN106802956B (en) | Movie recommendation method based on weighted heterogeneous information network | |
| Taleb et al. | Big data quality: A quality dimensions evaluation | |
| Park et al. | Reversed CF: A fast collaborative filtering algorithm using a k-nearest neighbor graph | |
| CN109933721B (en) | Interpretable recommendation method integrating user implicit article preference and implicit trust | |
| WO2020147720A1 (en) | Information recommendation method and device, and storage medium | |
| Liu et al. | Social temporal collaborative ranking for context aware movie recommendation | |
| US10904599B2 (en) | Predicting digital personas for digital-content recommendations using a machine-learning-based persona classifier | |
| US20180181667A1 (en) | System and method to model recognition statistics of data objects in a business database | |
| KR20160064447A (en) | A recommendation method for new users by using preference prediction based on collaborative filtering algorithm | |
| KR20160064448A (en) | A recommendation method for items by using preference prediction of their similar group | |
| Zhang et al. | A dynamic trust based two-layer neighbor selection scheme towards online recommender systems | |
| Liu et al. | Sequence-based trust in collaborative filtering for document recommendation | |
| CN104063555B (en) | The user model modeling method intelligently distributed towards remote sensing information | |
| US11443237B1 (en) | Centralized platform for enhanced automated machine learning using disparate datasets | |
| Chen et al. | A fuzzy matrix factor recommendation method with forgetting function and user features | |
| Deng | Leveraging consumer behaviors for product recommendation: an approach based on heterogeneous network | |
| Ali et al. | Dynamic context management in context-aware recommender systems | |
| Huang et al. | Efficient classification of distribution-based data for Internet of Things | |
| Luo et al. | A hybrid user profile model for personalized recommender system with linked open data | |
| Chen et al. | Trust-based collaborative filtering algorithm in social network | |
| Buvaneswari et al. | Comprehensive personalized recommendation technologies | |
| US9678982B2 (en) | Accessibility advisement system for digital assets | |
| Liu et al. | An online activity recommendation approach based on the dynamic adjustment of recommendation lists | |
| Dixit et al. | Proposed similarity measure using Bhattacharyya coefficient for context aware recommender system |
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: 14908869 Country of ref document: EP Kind code of ref document: A1 |
|
| NENP | Non-entry into the national phase |
Ref country code: DE |
|
| 122 | Ep: pct application non-entry in european phase |
Ref document number: 14908869 Country of ref document: EP Kind code of ref document: A1 |