技术领域Technical field
本发明涉及信息推荐的技术领域，具体涉及一种基于协同过滤的社交网络推荐装置及方法。The present invention relates to the technical field of information recommendation, and in particular to a social network recommendation device and method based on collaborative filtering.
背景技术Background technique
概述互联网信息的迅速增长产生了海量的数据,用户往往要耗费大量的时间和精力,才能找到自己感兴趣的信息。低效率的互联网信息检索技术已经成为阻碍用户有效利用信息的瓶颈,推荐技术就是在这样的背景下产生的。推荐技术可以有效解决信息过载的问题,从海量数据中检索到对用户有所帮助的内容。目前的推荐技术在同一时刻一般都是为单个用户进行推荐,但现实应用中可能经常需要为某一群体进行推荐,比如一个旅游团要旅游的目的地、一次聚会的用餐地点以及一个家庭的观影计划等。群体成员间的兴趣爱好具有很大的差异性,所以传统的推荐技术已经难以适用于群体推荐的要求,研究基于社交网络的群体推荐技术就具有重要的现实意义。Overview The rapid growth of Internet information has produced massive amounts of data, and users often have to spend a lot of time and energy to find the information they are interested in. Inefficient Internet information retrieval technology has become a bottleneck that hinders users from effectively utilizing information. It is against this background that recommendation technology was born. Recommendation technology can effectively solve the problem of information overload and retrieve content that is helpful to users from massive data. Current recommendation technology generally recommends for a single user at the same time, but in real applications it may often be necessary to recommend a certain group, such as the destination for a tour group, the dining place for a party, and the views of a family. Film project, etc. Interests and hobbies among group members are very different, so traditional recommendation technology has been difficult to meet the requirements of group recommendation. Research on group recommendation technology based on social networks has important practical significance.
推荐装置及社交网络群体推荐装置的研究涉及到个人推荐装置、社交网络以及群体决策等技术领域。现存的推荐装置多数为个人推荐装置,即推荐行为旨在为单个用户提供服务,其最常用的推荐算法为协同过滤推荐算法和基于内容的推荐算法；协同过滤算法借鉴与被推荐用户具有相似兴趣的用户的行为进行推荐,基于内容的推荐算法分析被推荐用户访问的历史内容,并利用不同内容的相似程度来进行推荐。Research on recommendation devices and social network group recommendation devices involves technical fields such as personal recommendation devices, social networks, and group decision-making. Most of the existing recommendation devices are personal recommendation devices, that is, the recommendation behavior is designed to provide services for a single user. The most commonly used recommendation algorithms are collaborative filtering recommendation algorithms and content-based recommendation algorithms; collaborative filtering algorithms draw on users who have similar interests as the recommended users. The content-based recommendation algorithm analyzes the historical content visited by the recommended user and uses the similarity of different content to make recommendations.
发明内容Contents of the invention
本发明的第一个目的是为了解决现有技术中的上述缺陷，提供一种基于协同过滤的社交网络推荐装置。The first object of the present invention is to solve the above-mentioned defects in the existing technology and provide a social network recommendation device based on collaborative filtering.
本发明的另一个目的是为了解决现有技术中的上述缺陷，提供一种基于协同过滤的社交网络推荐方法。Another object of the present invention is to provide a social network recommendation method based on collaborative filtering to solve the above-mentioned defects in the existing technology.
本发明的第一个目的可以通过采取如下技术方案达到：The first object of the present invention can be achieved by adopting the following technical solutions:
一种基于协同过滤的社交网络推荐装置，所述推荐装置包括依次连接的启动模块、过滤模块、推荐模块和排序模块，A social network recommendation device based on collaborative filtering, the recommendation device includes a startup module, a filtering module, a recommendation module and a sorting module connected in sequence,
其中，所述启动模块用于对推荐社交网络用户的行为数据和个人属性进行初始化定义，将用户的行为分类并设置初始搜索引擎的条件，向推荐引擎发起推荐请求，并将初始化的数据集发送给推荐引擎；Among them, the startup module is used to initialize and define the behavioral data and personal attributes of recommended social network users, classify the user's behavior and set the conditions of the initial search engine, initiate a recommendation request to the recommendation engine, and send the initialized data set to recommendation engines;
所述过滤模块通过社交网络中设置的基础搜索条件，由基础搜索引擎过滤掉不符合要求的用户，形成可推荐候选集A；The filtering module uses the basic search conditions set in the social network to filter out users who do not meet the requirements by the basic search engine to form a recommendable candidate set A;
所述推荐模块将可推荐候选集A根据用户历史行为和数据进行筛选，得出基本的用户感兴趣集，并基于用户的协同过滤思想，得到各个用户的可推荐结果集B；The recommendation module filters the recommendable candidate set A based on the user's historical behavior and data to obtain the basic user interest set, and based on the user's collaborative filtering idea, obtains the recommendable result set B for each user;
所述排序模块根据可推荐结果集B，对社交网络用户的行为数据和个人属性进行划分，将社交网络用户感兴趣集进行排序，得到初步推荐列表C。The sorting module divides the behavioral data and personal attributes of social network users according to the recommendable result set B, sorts the social network users' interest sets, and obtains a preliminary recommendation list C.
进一步地，用户的初始个人属性由用户在注册时填写，推荐引擎可以根据用户填写的个人属性值，为初始用户进行特征划分。Furthermore, the user's initial personal attributes are filled in by the user when registering, and the recommendation engine can classify the characteristics of the initial user based on the personal attribute values filled in by the user.
进一步地，当过滤筛选后的可推荐候选集A中人数不足最低计算人数时，所述启动模块中推荐引擎将发送指令，由社交网络扩大筛选范围或者增加筛选人数。Further, when the number of people in the filtered recommendable candidate set A is less than the minimum calculated number of people, the recommendation engine in the startup module will send an instruction to expand the screening scope or increase the number of people to be screened by the social network.
本发明的另一个目的可以通过采取如下技术方案达到：Another object of the present invention can be achieved by adopting the following technical solutions:
一种基于协同过滤的社交网络推荐方法，所述推荐方法包括下列步骤：A social network recommendation method based on collaborative filtering, the recommendation method includes the following steps:
S1、启动模块通过业务方对推荐社交网络用户的行为数据和个人属性进行初始化定义，将用户的行为数据分类并设置初始搜索引擎的条件，完成初始化设定和推荐引擎接入后，向推荐引擎发起推荐请求，并将初始化的数据集发送给推荐引擎；S1. The startup module initializes and defines the behavioral data and personal attributes of recommended social network users through the business party, classifies the user's behavioral data and sets the conditions for the initial search engine. After completing the initialization settings and accessing the recommendation engine, it reports to the recommendation engine Initiate a recommendation request and send the initialized data set to the recommendation engine;
S2、过滤模块通过推荐引擎根据业务方的初始搜索引擎的条件，将满足搜索条件的数据，汇总形成初步可推荐候选集，对初步可推荐候选集通过相似度判定得到相似用户集，并基于用户的协同过滤思想，进行数据筛选得到可推荐候选集A；S2. The filtering module uses the recommendation engine to summarize the data that meets the search conditions according to the conditions of the business party's initial search engine to form a preliminary recommendable candidate set. It determines the similarity of the preliminary recommendable candidate set to obtain a similar user set, and based on the user Based on the collaborative filtering idea, the data is filtered to obtain the recommendable candidate set A;
S3、推荐模块将可推荐候选集A根据推荐社交网络用户历史行为和数据进行筛选，得出基本的用户感兴趣集，并基于用户的协同过滤思想，得到各个推荐社交网络用户的可推荐结果集B；S3. The recommendation module filters the recommendable candidate set A based on the historical behavior and data of recommended social network users to obtain the basic user interest set, and based on the user's collaborative filtering idea, obtains the recommendable result set for each recommended social network user. B;
S4、排序模块根据推荐社交网络用户的个人属性，对个人属性的特征值进行划分，根据用户的感兴趣集中各个特征值所占的比例，得到可推荐结果集B的用户中每个属性特征值所占的权重，并根据各个最显著的特征集去得到用户感知最敏感的属性，根据感知最敏感的属性的不同参考权重，将用户的感兴趣集进行排序，得到初步推荐列表C。S4. The sorting module divides the characteristic values of the personal attributes according to the personal attributes of the recommended social network users, and obtains the characteristic values of each attribute in the users who can recommend the result set B according to the proportion of each characteristic value in the user's interest set. According to the weight of each most significant feature set, the user's most sensitive attributes are obtained. According to the different reference weights of the most sensitive attributes, the user's interest sets are sorted to obtain a preliminary recommendation list C.
进一步地，所述步骤S2具体过程如下：Further, the specific process of step S2 is as follows:
S201、社交网络给用户设定一个初始筛选条件，由用户进行选择或直接根据用户的历史行为进行设定。S201. The social network sets an initial filtering condition for the user, which is selected by the user or set directly based on the user's historical behavior.
S202、社交网络将初始筛选条件和初始化数据集发送给推荐引擎，推荐引擎根据这些条件进行筛选，形成可推荐候选集A。S202. The social network sends the initial screening conditions and initialization data set to the recommendation engine, and the recommendation engine performs screening based on these conditions to form a recommendable candidate set A.
进一步地，所述步骤S3具体过程如下：Further, the specific process of step S3 is as follows:
S301、将用户的行为分为T1～TK共K类，并对这K类行为分别进行权重赋值w1～wk，根据不同的用户行为区分为正面、负面以及高、中、低六个维度，赋值向量w的取值为w＝【-2,-1,0,1,2,3】；S301. Divide user behaviors into K categories T1 to TK in total, and assign weights w1 to wk to these K categories of behaviors respectively, and classify them into positive, negative, high, medium, and low according to different user behaviors. dimensions, the value of the assignment vector w is w=[-2,-1,0,1,2,3];
S302、获取用户对社交网络用户的行为操作累加值得到用户对社交网络用户的喜好度H＝∑w；S302. Obtain the accumulated value of the user's behavioral operations on the social network user to obtain the user's preference for the social network user H=∑w;
S303、通过不同用户对各个社交网络用户的喜好度H，利用欧几里得距离S303. Use the Euclidean distance based on the preference H of different users for each social network user.
计算得到用户之间的相似度：Calculate the similarity between users:
当两个用户之间的相似度sim(x,y)>k时，其中k由业务方决定，即认为两者相似从而得到相似用户集，并基于用户的协同过滤思想，得到各个用户的可推荐结果集B。When the similarity between two users is sim(x,y)>k, where k is determined by the business side, that is, the two users are considered similar to obtain a similar user set, and based on the user's collaborative filtering idea, the trustworthiness of each user is obtained. Recommended result set B.
进一步地，所述步骤S4具体过程如下：Further, the specific process of step S4 is as follows:
S401、根据可推荐结果集B，对推荐社交网络用户的个人属性和特征值进行划分，设推荐社交网络用户的个人属性向量为：S401. According to the recommendable result set B, divide the personal attributes and characteristic values of the recommended social network users. Suppose the personal attribute vector of the recommended social network users is:
属性Si的特征值向量为：The eigenvalue vector of attribute Si is:
S402、通过各个个人属性的区分度向量和候选推荐集构造属性特征矩阵；S402. Construct an attribute feature matrix through the distinction vector of each personal attribute and the candidate recommendation set;
对用户A而言，候选推荐集QT中某一社交网络用户的个人属性Si的特征值vk在所占的比例为/>则在候选推荐集QT中，设该社交网络用户的个人属性的特征值vk所占的权重：For user A, the characteristic value vk of the personal attributeSi of a certain social network user in the candidate recommendation set QT is at The proportion is/> Then in the candidate recommendation set QT , let the weight of the characteristic value vk of the social network user's personal attributes be:
另，当(i，k取任意可取值)Also, when (i, k can take any value)
则认为个人属性Sx的区分度最强；各个属性的区分度取：Then it is considered that the distinction of personal attribute Sx is the strongest; the distinction of each attribute is:
所以，当给出QA后，可得索性区分度向量：Therefore, when QA is given, the simple discrimination vector can be obtained:
得到QT后，可得社交网络用户的属性特征矩阵：After obtaining QT , the attribute characteristic matrix of social network users can be obtained:
S403、根据属性特征矩阵和属性区分度向量，可以得到候选推荐集QT的各社交网络用户的推荐分数向量：S403. According to the attribute feature matrix and attribute distinction vector, the recommendation score vector of each social network user of the candidate recommendation set QT can be obtained:
S404、根据得到的推荐分数向量对可推荐结果集B的排序，确定初步推荐列表C。S404. Sort the recommendable result set B according to the obtained recommendation score vector, and determine the preliminary recommendation list C.
进一步地，所述推荐社交网络用户的个人属性及特征值的定义的方式为【属性-值】键值对。Further, the recommended way of defining the personal attributes and characteristic values of social network users is [attribute-value] key-value pairs.
进一步地，当用户对社交网络用户的喜好度H>3时，则认为用户对该物品感兴趣。Furthermore, when the user's preference for social network users H>3, the user is considered to be interested in the item.
本发明相对于现有技术具有如下的优点及效果：Compared with the existing technology, the present invention has the following advantages and effects:
本发明公开了一种基于协同过滤的社交网络推荐装置及方法，相较于传统的基于人口统计学和基于内容的推荐装置相比，该社交网络推荐装置强调了用户之间的差异性，可针对用户的历史数据不断学习并革新推荐引擎，具有更强的鲁棒性。并且，利用协同过滤思想，更符合用户在真实社交场景中的交友情景，推荐的结果也会趋向于精确化。The invention discloses a social network recommendation device and method based on collaborative filtering. Compared with traditional demography-based and content-based recommendation devices, the social network recommendation device emphasizes the differences between users and can Continuously learn and innovate the recommendation engine based on users' historical data, making it more robust. Moreover, the use of collaborative filtering ideas is more in line with users' friendship situations in real social scenes, and the recommended results will also tend to be more accurate.
附图说明Description of the drawings
图1是本发明公开的基于协同过滤的社交网络推荐装置的组成框图；Figure 1 is a block diagram of a social network recommendation device based on collaborative filtering disclosed in the present invention;
图2是本发明公开的基于协同过滤的社交网络推荐方法的流程图。Figure 2 is a flow chart of the social network recommendation method based on collaborative filtering disclosed in the present invention.
具体实施方式Detailed ways
为使本发明实施例的目的、技术方案和优点更加清楚，下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例是本发明一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有作出创造性劳动前提下所获得的所有其他实施例，都属于本发明保护的范围。In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments These are some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts fall within the scope of protection of the present invention.
实施例Example
如附图1所示，本实施例公开了一种基于协调过滤的社交网络推荐装置，该社交网络推荐装置借鉴协同过滤思想和基于内容的推荐思想，设计一套社交网络交友的推荐引擎。As shown in Figure 1, this embodiment discloses a social network recommendation device based on coordinated filtering. The social network recommendation device draws on collaborative filtering ideas and content-based recommendation ideas to design a set of social network friend recommendation engines.
该社交网络推荐装置具体包括依次连接的启动模块、过滤模块、推荐模块和排序模块，The social network recommendation device specifically includes a startup module, a filtering module, a recommendation module and a sorting module that are connected in sequence.
其中，所述启动模块用于对推荐社交网络用户的行为数据和个人属性进行初始化定义，将用户的行为分类并设置初始搜索引擎的条件，向推荐引擎发起推荐请求，并将初始化的数据集发送给推荐引擎；Among them, the startup module is used to initialize and define the behavioral data and personal attributes of recommended social network users, classify the user's behavior and set the conditions of the initial search engine, initiate a recommendation request to the recommendation engine, and send the initialized data set to recommendation engines;
具体实施方式中，所述启动模块的工作过程如下：In a specific implementation, the working process of the startup module is as follows:
业务方对推荐社交网络用户的行为数据和个人属性进行初始化定义，将用户的行为分类并设置初始搜索引擎的条件，完成初始化设定和推荐引擎接入后，向推荐引擎发起推荐请求，并将初始化的数据集发送给推荐引擎。The business party initializes and defines the behavioral data and personal attributes of the recommended social network users, classifies the user's behavior and sets the conditions for the initial search engine. After completing the initial settings and accessing the recommendation engine, it initiates a recommendation request to the recommendation engine, and The initialized data set is sent to the recommendation engine.
具体应用中，用户的初始个人属性由用户在注册时填写，推荐引擎可以根据用户填写的个人属性值，为初始用户进行特征划分。In specific applications, the user's initial personal attributes are filled in by the user when registering, and the recommendation engine can classify the characteristics of the initial user based on the personal attribute values filled in by the user.
其中，所述过滤模块通过社交网络中设置的基础搜索条件，由基础搜索引擎过滤掉不符合要求的用户，形成可推荐候选集A，实现精简推荐装置计算量的目的。Among them, the filtering module uses the basic search conditions set in the social network to filter out users who do not meet the requirements by the basic search engine to form a recommendable candidate set A, thereby achieving the purpose of streamlining the calculation amount of the recommendation device.
主要工作原理如下：The main working principle is as follows:
1)社交网络给用户设定一个初始筛选条件，由用户进行选择或直接根据用户的历史行为进行设定。1) The social network sets an initial filtering condition for the user, which is selected by the user or set directly based on the user's historical behavior.
2)社交网络将初始筛选条件和初始化数据集发送给推荐引擎，推荐引擎根据这些条件进行筛选，形成可推荐候选集A。2) The social network sends the initial screening conditions and initialization data set to the recommendation engine, and the recommendation engine filters based on these conditions to form a set of recommendable candidates A.
3)当筛选后的人数不足最低计算人数时，推荐引擎将发送指令，由社交网络扩大筛选范围或者增加筛选人数。3) When the number of people after screening is less than the minimum number of people to be calculated, the recommendation engine will send instructions to expand the screening scope or increase the number of people to be screened by the social network.
其中，所述推荐模块将可推荐候选集A根据用户历史行为和数据进行筛选，得出基本的用户感兴趣集，并基于用户的协同过滤思想，得到各个用户的可推荐结果集B。Among them, the recommendation module filters the recommendable candidate set A according to the user's historical behavior and data to obtain the basic user interest set, and obtains the recommendable result set B for each user based on the user's collaborative filtering idea.
该模块为推荐装置的核心模块，主要目的是将筛选后的用户根据用户历史行为和数据进一步进行筛选，得出基本的用户感兴趣集。This module is the core module of the recommendation device. Its main purpose is to further filter the filtered users based on user historical behavior and data to obtain a basic user interest set.
主要工作原理如下：The main working principle is as follows:
1)将用户的行为进行分类，并对分类后的行为分别进行权重赋值。1) Classify the user's behavior and assign weights to the classified behaviors.
2)推荐模块获取用户对交友的行为操作累加值得到用户对社交网络用户的喜好度大于一定值时，则认为用户对该社交网络用户感兴趣；2) The recommendation module obtains the cumulative value of the user's behavior and operations on making friends and when the user's preference for the social network user is greater than a certain value, the user is considered to be interested in the social network user;
3)计算不同用户对各个其他社交网络用户感的喜好度，并通过不同用户的喜好度，利用欧几里得距离计算得到用户之间的相似度，当两个用户之间的相似度大于一定值时，既认为两者相似从而得到相似用户集，并基于用户的协同过滤思想，得到各个用户的可推荐结果集B。3) Calculate the preferences of different users for each other social network user, and use the preferences of different users to calculate the similarity between users using Euclidean distance. When the similarity between two users is greater than a certain value, it is considered that the two are similar to obtain a similar user set, and based on the user's collaborative filtering idea, a recommendable result set B for each user is obtained.
其中，所述排序模块根据可推荐结果集B，对社交网络用户的行为数据和个人属性进行划分，将社交网络用户感兴趣集进行排序，得到初步推荐列表C。Among them, the sorting module divides the behavioral data and personal attributes of the social network users according to the recommendable result set B, sorts the social network users' interest sets, and obtains the preliminary recommendation list C.
主要工作原理如下：The main working principle is as follows:
1)根据社交网络用户的个人属性，对个人属性的特征值进行划分，根据社交网络用户感兴趣集中各个特征值所占的比例，得到可推荐候选集A中用户的每个属性特征值所占的权重，并根据各个最显著的特征集去得到用户感知最敏感的属性，以此得出各个属性的不同参考权重。1) According to the personal attributes of social network users, the characteristic values of personal attributes are divided. According to the proportion of each characteristic value in the social network user's interest set, the proportion of each attribute characteristic value of the user in the recommendable candidate set A is obtained. weight, and obtain the most sensitive attributes perceived by users based on each of the most significant feature sets, thereby obtaining different reference weights for each attribute.
2)通过各个属性的区分度向量和候选推荐集的属性特征矩阵，得到候选推荐集的得分向量，根据得分向量对可推荐结果集B的排序，得到初步推荐列表C，即为最终的推荐结果。2) Through the distinction vector of each attribute and the attribute feature matrix of the candidate recommendation set, the score vector of the candidate recommendation set is obtained. According to the score vector, the recommendable result set B is sorted to obtain the preliminary recommendation list C, which is the final recommendation result. .
实施例二Embodiment 2
如附图2所示，本实施例公开了一种基于协调过滤的社交网络推荐方法，该社交网络推荐方法借鉴协同过滤思想和基于内容的推荐思想，设计一套社交网络交友的推荐引擎。As shown in Figure 2, this embodiment discloses a social network recommendation method based on coordinated filtering. The social network recommendation method draws on collaborative filtering ideas and content-based recommendation ideas to design a set of social network friend recommendation engines.
该社交网络推荐方法具体包括下列步骤：The social network recommendation method specifically includes the following steps:
S1、启动模块通过业务方对推荐社交网络用户的行为数据和个人属性进行初始化定义，定义的方式为【属性-值】键值对，例如【身高-170CM】；将用户的行为数据分类并设置初始搜索引擎的条件，完成初始化设定和推荐引擎接入后，向推荐引擎发起推荐请求，并将初始化的数据集发送给推荐引擎。S1. The startup module initializes and defines the behavioral data and personal attributes of recommended social network users through the business party. The definition method is [attribute-value] key-value pairs, such as [height-170CM]; the user's behavioral data is classified and set Initial search engine conditions, after completing the initial settings and recommendation engine access, initiate a recommendation request to the recommendation engine, and send the initialized data set to the recommendation engine.
具体应用中，用户的初始属性由用户在注册时填写，推荐引擎可以根据用户填写的属性值，为初始用户进行特征划分。In specific applications, the user's initial attributes are filled in by the user when registering, and the recommendation engine can classify the characteristics of the initial user based on the attribute values filled in by the user.
S2、过滤模块通过推荐引擎根据业务方的初始搜索引擎的条件，将满足搜索条件的数据，汇总形成初步可推荐候选集，对初步可推荐候选集通过相似度判定得到相似用户集，并基于用户的协同过滤思想，进行数据筛选得到可推荐候选集A；如果可推荐候选集A的人数不满足最低推荐人数要求，则向业务方请求扩大搜索条件。S2. The filtering module uses the recommendation engine to summarize the data that meets the search conditions according to the conditions of the business party's initial search engine to form a preliminary recommendable candidate set. It determines the similarity of the preliminary recommendable candidate set to obtain a similar user set, and based on the user Based on the collaborative filtering idea, the data is filtered to obtain the recommendable candidate set A; if the number of people who can recommend the candidate set A does not meet the minimum number of recommenders, request the business party to expand the search conditions.
该步骤具体过程如下：The specific process of this step is as follows:
S201、社交网络给用户设定一个初始筛选条件，由用户进行选择或直接根据用户的历史行为进行设定。S201. The social network sets an initial filtering condition for the user, which is selected by the user or set directly based on the user's historical behavior.
S202、社交网络将初始筛选条件和初始化数据集发送给推荐引擎，推荐引擎根据这些条件进行筛选，形成可推荐候选集A。S202. The social network sends the initial screening conditions and initialization data set to the recommendation engine, and the recommendation engine performs screening based on these conditions to form a recommendable candidate set A.
S3、推荐模块将可推荐候选集A根据推荐社交网络用户历史行为和数据进行筛选，得出基本的用户感兴趣集，并基于用户的协同过滤思想，得到各个用户的可推荐结果集B。S3. The recommendation module filters the recommendable candidate set A based on the historical behavior and data of recommended social network users to obtain the basic user interest set, and based on the user's collaborative filtering idea, obtains the recommendable result set B for each user.
该步骤S3的具体过程如下：The specific process of step S3 is as follows:
S301、将用户的行为分为T1～TK共K类，并对这K类行为分别进行权重赋值w1～wk，根据不同的用户行为区分为正面、负面以及高、中、低六个维度，赋值向量w的取值为w＝【-2,-1,0,1,2,3】；S301. Divide user behaviors into K categories T1 to TK in total, and assign weights w1 to wk to these K categories of behaviors respectively, and classify them into positive, negative, high, medium, and low according to different user behaviors. dimensions, the value of the assignment vector w is w=[-2,-1,0,1,2,3];
具体应用中，根据用户在线上的行为数据，定义不同行为的特征。In specific applications, the characteristics of different behaviors are defined based on users' online behavioral data.
S302、获取用户对社交网络用户的行为操作累加值得到用户对社交网络用户的喜好度H＝∑w，当H>3时，则认为用户对该社交网络用户感兴趣；S302. Obtain the accumulated value of the user's behavioral operations on the social network user to obtain the user's preference for the social network user H=∑w. When H>3, the user is considered to be interested in the social network user;
S303、通过不同用户对各个社交网络用户的喜好度H，利用欧几里得距离S303. Use the Euclidean distance based on the preference H of different users for each social network user.
计算得到用户之间的相似度：Calculate the similarity between users:
当两个用户之间的相似度sim(x,y)>k时，其中k由业务方决定，即认为两者相似从而得到相似用户集，并基于用户的协同过滤思想，得到各个用户的可推荐结果集B。When the similarity between two users is sim(x,y)>k, where k is determined by the business side, that is, the two users are considered similar to obtain a similar user set, and based on the user's collaborative filtering idea, the trustworthiness of each user is obtained. Recommended result set B.
S4、排序模块根据推荐社交网络用户的个人属性，对个人属性的特征值进行划分，根据用户的感兴趣集中各个特征值所占的比例，得到可推荐结果集B的用户中每个属性特征值所占的权重，并根据各个最显著的特征集去得到用户感知最敏感的属性，根据感知最敏感的属性的不同参考权重，将用户的感兴趣集进行排序，得到初步推荐列表C。S4. The sorting module divides the characteristic values of the personal attributes according to the personal attributes of the recommended social network users, and obtains the characteristic values of each attribute in the users who can recommend the result set B according to the proportion of each characteristic value in the user's interest set. According to the weight of each most significant feature set, the user's most sensitive attributes are obtained. According to the different reference weights of the most sensitive attributes, the user's interest sets are sorted to obtain a preliminary recommendation list C.
该步骤具体过程如下：The specific process of this step is as follows:
S401、根据可推荐结果集B，对推荐社交网络用户的个人属性和特征值进行划分，设推荐社交网络用户的个人属性向量为：S401. According to the recommendable result set B, divide the personal attributes and characteristic values of the recommended social network users. Suppose the personal attribute vector of the recommended social network users is:
属性Si的特征值向量为：The eigenvalue vector of attribute Si is:
S402、通过各个个人属性的区分度向量和候选推荐集构造属性特征矩阵；S402. Construct an attribute feature matrix through the distinction vector of each personal attribute and the candidate recommendation set;
对用户A而言，候选推荐集QT中某一社交网络用户的个人属性Si的特征值vk在所占的比例为/>则在候选推荐集QT中，设该社交网络用户的个人属性的特征值vk所占的权重：For user A, the characteristic value vk of the personal attributeSi of a certain social network user in the candidate recommendation set QT is at The proportion is/> Then in the candidate recommendation set QT , let the weight of the characteristic value vk of the social network user's personal attributes be:
另，当(i，k取任意可取值)Also, when (i, k can take any value)
则认为个人属性Sx的区分度最强；各个属性的区分度取：Then it is considered that the distinction of personal attribute Sx is the strongest; the distinction of each attribute is:
所以，当给出QA后，可得索性区分度向量：Therefore, when QA is given, the simple discrimination vector can be obtained:
得到QT后，可得社交网络用户的属性特征矩阵：After obtaining QT , the attribute characteristic matrix of social network users can be obtained:
S403、根据属性特征矩阵和属性区分度向量，可以得到候选推荐集QT的各社交网络用户的推荐分数向量：S403. According to the attribute feature matrix and attribute distinction vector, the recommendation score vector of each social network user of the candidate recommendation set QT can be obtained:
S404、根据得到的推荐分数向量对可推荐结果集B的排序，确定初步推荐列表C。S404. Sort the recommendable result set B according to the obtained recommendation score vector, and determine the preliminary recommendation list C.
上述实施例为本发明较佳的实施方式，但本发明的实施方式并不受上述实施例的限制，其他的任何未背离本发明的精神实质与原理下所作的改变、修饰、替代、组合、简化，均应为等效的置换方式，都包含在本发明的保护范围之内。The above embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above embodiments. Any other changes, modifications, substitutions, combinations, etc. may be made without departing from the spirit and principles of the present invention. All simplifications should be equivalent substitutions, and are all included in the protection scope of the present invention.
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710106305.4ACN106709076B (en) | 2017-02-27 | 2017-02-27 | Social network recommendation device and method based on collaborative filtering |
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710106305.4ACN106709076B (en) | 2017-02-27 | 2017-02-27 | Social network recommendation device and method based on collaborative filtering |
| Publication Number | Publication Date |
|---|---|
| CN106709076A CN106709076A (en) | 2017-05-24 |
| CN106709076Btrue CN106709076B (en) | 2023-09-29 |
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710106305.4AActiveCN106709076B (en) | 2017-02-27 | 2017-02-27 | Social network recommendation device and method based on collaborative filtering |
| Country | Link |
|---|---|
| CN (1) | CN106709076B (en) |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110020099A (en)* | 2017-08-21 | 2019-07-16 | 武汉斗鱼网络科技有限公司 | A kind of the user's recommended method and device of video friend-making |
| CN108090801A (en)* | 2017-11-29 | 2018-05-29 | 维沃移动通信有限公司 | Method of Commodity Recommendation, mobile terminal and server |
| CN108763314B (en) | 2018-04-26 | 2021-01-19 | 深圳市腾讯计算机系统有限公司 | Interest recommendation method, device, server and storage medium |
| CN109739768B (en)* | 2018-12-29 | 2021-03-30 | 深圳Tcl新技术有限公司 | Search engine evaluation method, device, equipment and readable storage medium |
| US11429682B2 (en)* | 2019-06-25 | 2022-08-30 | Sap Portals Israel Ltd. | Artificial crowd intelligence via networking recommendation engines |
| CN111797309B (en)* | 2020-06-19 | 2024-04-16 | 一汽奔腾轿车有限公司 | Vehicle-mounted intelligent recommendation device and method based on travel mode |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2011101527A1 (en)* | 2010-02-19 | 2011-08-25 | Osumus Recommendations Oy | Method for providing a recommendation to a user |
| CN103927347A (en)* | 2014-04-01 | 2014-07-16 | 复旦大学 | Collaborative filtering recommendation algorithm based on user behavior models and ant colony clustering |
| CN104317900A (en)* | 2014-10-24 | 2015-01-28 | 重庆邮电大学 | Multiattribute collaborative filtering recommendation method oriented to social network |
| KR101516329B1 (en)* | 2013-12-27 | 2015-05-06 | 충북대학교 산학협력단 | System and method for recommending group in social network environments |
| CN105095267A (en)* | 2014-05-09 | 2015-11-25 | 阿里巴巴集团控股有限公司 | User involving project recommendation method and apparatus |
| CN106327227A (en)* | 2015-06-19 | 2017-01-11 | 北京航天在线网络科技有限公司 | Information recommendation system and information recommendation method |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090164929A1 (en)* | 2007-12-20 | 2009-06-25 | Microsoft Corporation | Customizing Search Results |
| US20150286650A1 (en)* | 2014-04-03 | 2015-10-08 | Kurt Stump | Decision Making and Activity Recommendations Engine via Online Persona |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2011101527A1 (en)* | 2010-02-19 | 2011-08-25 | Osumus Recommendations Oy | Method for providing a recommendation to a user |
| KR101516329B1 (en)* | 2013-12-27 | 2015-05-06 | 충북대학교 산학협력단 | System and method for recommending group in social network environments |
| CN103927347A (en)* | 2014-04-01 | 2014-07-16 | 复旦大学 | Collaborative filtering recommendation algorithm based on user behavior models and ant colony clustering |
| CN105095267A (en)* | 2014-05-09 | 2015-11-25 | 阿里巴巴集团控股有限公司 | User involving project recommendation method and apparatus |
| CN104317900A (en)* | 2014-10-24 | 2015-01-28 | 重庆邮电大学 | Multiattribute collaborative filtering recommendation method oriented to social network |
| CN106327227A (en)* | 2015-06-19 | 2017-01-11 | 北京航天在线网络科技有限公司 | Information recommendation system and information recommendation method |
| Title |
|---|
| 基于产品属性及用户偏好的个性化产品推荐方法;鲁辉;张太华;何二宝;徐卫平;;贵州师范大学学报(自然科学版)(第01期);87-92* |
| 基于双边兴趣的社交网好友推荐方法研究;何静等;《计算机工程与应用》;20151231(第06期);108-113页* |
| Publication number | Publication date |
|---|---|
| CN106709076A (en) | 2017-05-24 |
| Publication | Publication Date | Title |
|---|---|---|
| CN106709076B (en) | Social network recommendation device and method based on collaborative filtering | |
| CN110598130B (en) | A movie recommendation method integrating heterogeneous information network and deep learning | |
| CN106126669B (en) | Tag-based user collaborative filtering content recommendation method and device | |
| CN104462385B (en) | A kind of film personalization similarity calculating method based on user interest model | |
| CN108154425B (en) | Offline merchant recommendation method combining social network and location | |
| WO2019205795A1 (en) | Interest recommendation method, computer device, and storage medium | |
| CN105138653B (en) | It is a kind of that method and its recommendation apparatus are recommended based on typical degree and the topic of difficulty | |
| CN107526850A (en) | Social networks friend recommendation method based on multiple personality feature mixed architecture | |
| CN113836393B (en) | Cold start recommendation method based on preference self-adaptive meta-learning | |
| CN106484876A (en) | A kind of based on typical degree and the collaborative filtering recommending method of trust network | |
| CN109241454A (en) | A kind of point of interest recommended method merging social networks and picture material | |
| CN107122447A (en) | The network searching system and control method of a kind of multi-data source fusion based on preference | |
| CN106951436B (en) | Large-scale online recommendation method based on mobile situation | |
| CN106952130A (en) | A general item recommendation method based on collaborative filtering | |
| CN107346333B (en) | Online social network friend recommendation method and system based on link prediction | |
| CN104778237A (en) | Individual recommending method and system based on key users | |
| CN107103499A (en) | Method is recommended in a kind of cross-site cold start-up based on deep learning | |
| US20240248944A1 (en) | Systems and methods for implementing session cookies for content selection | |
| CN107896153A (en) | A kind of flow package recommendation method and device based on mobile subscriber's internet behavior | |
| CN108241619A (en) | A Recommendation Method Based on Multiple Interests of Users | |
| WO2024114034A1 (en) | Content recommendation method and apparatus, device, medium, and program product | |
| CN113449200A (en) | Article recommendation method and device and computer storage medium | |
| WO2023231542A1 (en) | Representation information determination method and apparatus, and device and storage medium | |
| CN103577400A (en) | Location information providing method and system | |
| CN112380451A (en) | Favorite content recommendation method based on big data |
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |