移动社交网络信息过滤及推荐系统研究

发布时间：2018-03-29 03:11

本文选题：社交网络　切入点：信息过滤　出处：《南京邮电大学》2015年硕士论文

【摘要】：随着移动互联网的快速发展,移动社交网络日渐成为我们生活中不可缺的一部分,移动社交网络具有的广大的用户群体、迅捷、开放式的信息传播方式特性对商品营销而言极具吸引力。但是现有的移动社交网络缺乏正规友好的商品宣传机制和虚假伪劣商品信息甄别机制,从而导致移动社交网络中各种广告宣传无孔不入,广告信息真假难辨,严重影响正常用户的社交体感。为了解决这一问题,本文从两个方向进行解决。一方面,为了解决垃圾广告信息影响正常用户社交的问题,文本研究了垃圾信息过滤相关现有技术,其中基于机器学习分类的过滤方法正确率高、成本低,特别是其中的SVM分类方法。但SVM分类方法训练时间过长,不能灵活应对数据集的变化,针对此问题本文提出了一种改进的SVM增量学习算法,相比于传统的SVM增量学习算法,在保持准确率的同时,节约学习时间,提升学习效率。应用于垃圾广告信息过滤系统之中,获得较好的过滤效果。另一方面,通过研究社交网络中商品推荐现有的算法,对现有推荐方法,结合本课题实际需求,重点研究了社交网络信息对于商品推荐的积极意义,提出一种基于社交网络的混合推荐系统。以用户社会关系网络为基础,挖掘出与被推荐用户兴趣相似度最高的N个用户。与此同时,考虑社交网络的用户消息中包含的用户可能的商品需求,对用户的消息进行挖掘,将挖掘出的信息应用到推荐之中,获得了较好的推荐效果和用户满意度。最后,展示了本文提出的移动社交网络信息过滤及推荐系统在“友信”系统中的实现以及取得的效果,并分别针对过滤算法和推荐算法对算法性能进行仿真分析。测试结果表明本文提出的改进的SVM增量学习算法在垃圾信息过滤系统中取得了很好的过滤效果,基于社交网络的混合推荐方法也在契合课题系统需求的基础上获得了较好的推荐效果。
[Abstract]:With the rapid development of mobile Internet, mobile social network has become an indispensable part of our life. The characteristics of open information dissemination are very attractive to commodity marketing. However, the existing mobile social networks lack of formal and friendly commodity promotion mechanism and the identification mechanism of fake and inferior commodity information. As a result, all kinds of advertising in mobile social networks are ubiquitous, the advertising information is hard to distinguish, and the social sense of normal users is seriously affected. In order to solve this problem, this paper solves this problem from two directions. In order to solve the problem that spam information affects normal users' social interaction, the text studies the existing technologies of spam filtering, in which the filtering method based on machine learning classification has high accuracy and low cost. Especially the SVM classification method, but the training time of SVM classification method is too long, so it is not flexible to deal with the change of data set. In this paper, an improved SVM incremental learning algorithm is proposed, which is compared with the traditional SVM incremental learning algorithm. At the same time, it can save learning time and improve learning efficiency. It can be applied to spam information filtering system to obtain better filtering effect. On the other hand, by studying the existing algorithms of commodity recommendation in social networks, In this paper, the positive significance of social network information for commodity recommendation is mainly studied, and a hybrid recommendation system based on social network is proposed, which is based on user social network. At the same time, considering the possible commodity requirements of users included in the user messages of social networks, the users' messages are mined. The information extracted is applied to the recommendation system, and good recommendation effect and user satisfaction are obtained. Finally, the realization and effect of the mobile social network information filtering and recommendation system proposed in this paper in the "Friends letter" system are shown. The simulation results show that the improved SVM incremental learning algorithm has achieved a good filtering effect in the spam filtering system. The hybrid recommendation method based on social network also meets the requirements of the project system and obtains a good recommendation effect.
【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP391.3

【相似文献】