面向多媒体社交网络的情境分析与意图发现机制研究及应用

发布时间：2018-03-28 04:03

本文选题：多媒体社交网络　切入点：情境分析　出处：《河南科技大学》2017年硕士论文

【摘要】：随着多媒体社交网络(Multimedia Social Networks,简称MSNs)的快速发展,用户和数字内容的“爆炸式”增长,方便了用户对数字内容的访问,同时也增加了大量的用户与用户,用户与系统间的交互。针对这种复杂的交互如何给用户提供及时快捷的个性化服务成为多媒体社交网络研究的一个挑战。因此,本课题的研究在情境感知理论的基础上发现用户在多媒体社交网络中的频繁的行为模式,并根据用户的这种行为模式设计实现一种基于情境感知的多媒体内容推荐方法,使用户在海量的数据中发现可能感兴趣的内容。主要研究内容及创新之处如下:1.基于情境感知理论建立一种多媒体社交网络中用户频繁行为模式发现的方法。在多媒体社交网络环境中,大量用户可能在不同的群组中,在相应的群组中用户有不同的角色,用户在群组中的角色可能使用户产生不同的意图,所以在社交媒体下,首先结合社交媒体特有的属性在Carl K.Chang教授Situ理论的基础上建立了SocialSitu理论,对框架中的元素进行定义;其次,基于SocialSitu框架并结合改进的GSP(Generalized Sequential Pattern,广义序列模式)算法设计了多媒体社交网络中的用户行为模式发现方法;最后,通过该方法分析用户的历史SocialSitu(t),可以得出用户在不同意图下的行为模式,根据该用户的不同意图下的行为模式可以预测用户当前行为序列下的意图。2.基于用户的行为模式设计多媒体社交网络中的多媒体内容推荐算法。当前现有的推荐系统在用户对浏览的内容打分或进行其他操作之前,并不知道用户对当前浏览内容的喜好程度,并且用户的喜好可能随着用户所处的环境以及用户的身份随时发生变化,通常在多媒体社交网络中,用户有自己的评分习惯,或用户的评分可能很随意。因此传统的推荐算法仅利用相似用户对内容的评分来预测目标用户的评分是不准确的,但是用户对多媒体内容的操作行为却能真实的反应用户对音视频的喜好程度。所以为了及时给用户推荐可能感兴趣的内容,本课题的研究在SocialSitu理论的基础上提出了一种基于用户情境感知的多媒体内容推荐算法,根据目标用户的近邻集合对浏览内容的历史操作行为及评分记录,以及目标用户浏览多媒体内容时的意图预测分析其对当前内容的喜好程度,对目标用户推荐潜在的可能感兴趣的内容,通过实验分析,该推荐算法在准确率和召回率方面有很大提高。在多媒体社交网络环境下给用户提供个性化的服务已经成为当前服务系统的趋势,本课题在多媒体社交网络环境下,针对用户的历史行为数据分析挖掘用户的行为模式,从而根据这些模式分析用户潜在的喜好。本文的主要贡献在于:(1)扩展和丰富了Situ理论;(2)通过序列化算法得出用户不同意图下的行为序列模式,并根据某个用户在各种意图下的行为序列模式预测该用户在当前行为序列下的意图;(3)设计并实现了基于情境分析的多媒体内容推荐算法,可以及时的感知用户潜在的喜好,给用户提供及时的个性化推荐。
[Abstract]:Along with the multimedia social networks (Multimedia Social Networks, referred to as MSNs) the rapid development of users and digital content "explosive" growth, convenient for users to access the digital content, but also increase the number of users and users, interaction between users and system. In view of this complex interaction to provide personalized service timely and quickly become a challenge of multimedia social networks. Therefore, this research found that users in the multimedia social networks frequent behavior model based on context awareness theory, and according to the design of the user behavior implementation of a multimedia content recommendation method based on context awareness, users may find. Interested in the content in the massive data. The main research content and innovation are as follows: 1. the establishment of a social network based on the theory of context aware multimedia Method of network user frequent behavior patterns found in multimedia social network environment, a large number of users in different groups, in the corresponding group of users have different roles in user roles in the group of users may have different intentions, so in the social media, based on Carl K.Chang Situ the theory is established on the basis of SocialSitu theory of social media properties peculiar to the definition of the elements in the framework; secondly, combined with the SocialSitu framework based on improved GSP (Generalized Sequential Pattern, generalized sequential pattern) algorithm to design the model of user behavior of multimedia in social network discovery method; finally, through the analysis of users the history of SocialSitu (T), can be obtained in different intentions under the user behavior model, according to the user's intention of different patterns of behavior can predict the user when The intent of the.2. sequence before the behavior of user behavior patterns based on the design of multimedia content of the multimedia social network recommendation algorithm. The current recommendation system before the user to browse the contents of other operations or the user does not know the score, preference current browsing content, and user preferences may with user environment and the identity of the user to change at any time, usually in the multimedia social networks, users have their own scoring habits, or the user may score is random. So the traditional recommendation algorithm only uses the similar user score on the content to predict the target user's score is not accurate, but the operation behavior of users of multimedia content can like the reaction degree of real users of audio and video. So in order to recommend users may be interested in the contents of the research in Soci On the basis of alSitu theory proposes a recommendation algorithm based on user context aware multimedia content, according to the target user to browse the contents of the neighbor sets of historical operation behavior and scoring records, and intent to target users to browse multimedia content analysis and prediction of its preference for the former when the content, the target user recommended potential sense interested in the content, through experimental analysis, the recommendation algorithm is greatly improved in terms of accuracy and recall rate. In the multimedia social network environment to provide users with personalized service service system has become the current trend, this topic in the multimedia social network environment, according to the history of user behavior data mining and analysis of user behavior patterns thus, the potential user preferences according to the model analysis. The main contribution of this paper is: (1) extended and enriched the theory of Situ; (2) by The serialization algorithm behavior patterns of different user intent, and according to the behavior patterns of a user in a variety of intention prediction under the user in the current sequence of behavior intention; (3) the design and implementation of multimedia content recommendation algorithm based on scenario analysis, timely perception of potential user preferences, to provide users with timely personalized recommendation.

【学位授予单位】：河南科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.3;TP393.09

【相似文献】