基于上下文感知的个性化信息服务系统的研究与设计

发布时间：2018-11-18 08:59

【摘要】：随着信息技术和互联网技术的发展,人们渐渐地从信息匮乏过渡到了信息过载的时代,为了解决信息过载的问题,最具代表性的解决方案是分类目录和搜索引擎,但是这两种解决方案还不能完全满足人们的需求,于是个性化信息服务系统应运而生,该系统通过建立用户的兴趣模型为用户提供个性化的服务,其核心是推荐引擎。然而现有的个性化信息服务系统方面的研究并没有考虑到上下文的信息：比如时间、地理位置、同伴等。本论文针对传统的推荐算法的缺点提出了基于上下文的推荐算法,设计并实现了一个基于上下文的音乐推荐系统。本论文的主要工作如下：首先介绍了传统的协同过滤算法中应用最广的基于用户的协同过滤算法和基于物品的协同过滤算法的实现步骤,同时分析了这两种传统的协同过滤算法的优缺点。接着在分析了传统的协同过滤算法的优缺点的基础上,提出了基于上下文信息的推荐算法。首先从上下文的定义出发,然后阐述上下文的获取方式、上下文信息的建模方法,最后将基于上下文的推荐算法分为三个方式：上下文预过滤、上下文后过滤和上下文建模,并分别给出了相应的算法。然后用python语言实现了传统的推荐算法和三种基于上下文的推荐算法,并计算出每种算法的准确率、召回率、覆盖率和推荐物品的流行度,详细地分析比较了三种基于上下文的推荐算法和传统的推荐算法在准确度和挖掘长尾物品能力上的优劣性。最后设计并实现了一个基于上下文的音乐推荐系统,详细介绍了基于上下文音乐推荐系统的架构及其用户特征向量模块、特征-物品相关推荐模块和推荐列表过滤模块的工作原理,阐述了如何用wamp的方式开发基于上下文的音乐推荐系统,并对基于上下文的音乐推荐系统进行了测试。论文末尾对个性化信息服务系统的应用前景进行了总结与展望。
[Abstract]:With the development of information technology and Internet technology, people gradually transition from the lack of information to the era of information overload. In order to solve the problem of information overload, the most representative solution is classified catalogue and search engine. However, these two solutions can not fully meet the needs of people, so personalized information service system emerges as the times require. The system provides personalized services to users by building user interest model, the core of which is recommendation engine. However, the existing research on personalized information service system does not take into account the context information, such as time, geographical location, peer, and so on. Aiming at the shortcomings of traditional recommendation algorithms, this paper proposes a context-based recommendation algorithm, and designs and implements a context-based music recommendation system. The main work of this paper is as follows: firstly, the steps of the most widely used collaborative filtering algorithms based on users and articles based on collaborative filtering are introduced. At the same time, the advantages and disadvantages of these two traditional collaborative filtering algorithms are analyzed. Then, based on the analysis of the advantages and disadvantages of the traditional collaborative filtering algorithm, a context-based recommendation algorithm is proposed. Firstly, the definition of context is introduced, then the way to obtain context and the modeling method of context information are expounded. Finally, the recommendation algorithm based on context is divided into three ways: context pre-filtering, post-context filtering and context modeling. The corresponding algorithms are given respectively. Then the traditional recommendation algorithm and three context-based recommendation algorithms are implemented with python language, and the accuracy, recall, coverage and popularity of each algorithm are calculated. The advantages and disadvantages of three context-based recommendation algorithms and traditional recommendation algorithms in accuracy and ability to mine long-tailed items are analyzed and compared in detail. Finally, a context-based music recommendation system is designed and implemented. The architecture of the context-based music recommendation system and its user feature vector module are introduced in detail. The working principle of feature-item related recommendation module and recommendation list filtering module is discussed, and how to develop context-based music recommendation system with wamp is described, and the context-based music recommendation system is tested. At the end of the paper, the application prospect of personalized information service system is summarized and prospected.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP391.3

【参考文献】