【作者】 李明；

【导师】 孙天昊；

【作者基本信息】 重庆大学 ， 计算机技术（专业学位）， 2015， 硕士

【摘要】 当今,网络中充斥着各种各样的信息,信息过载问题日益严重。对于普通的网络用户而言,很容易被复杂的信息所干扰,使得他们需要花费成倍的时间以及精力去寻找自己需要的信息。而推荐系统为信息过载问题提供了一种有效的解决方法,它能够针对用户的特点为其提供个性化的推荐服务。推荐系统根据不同的应用需求使用的推荐算法不同,其中协同过滤推荐是目前最为流行并且实用的一种推荐技术。在本文中对现有基于信任的协同过滤推荐算法进行研究和改进,最后设计并实现了一个结合多算法电影推荐系统。本文的主要工作包括:①对推荐系统进行深入研究,分析其概念构成、评估标准等。对几种常用的推荐算法的流程、特点进行重点分析。②传统的协同过滤推荐存在数据稀疏性问题,这在很大程度上影响了推荐结果的准确性。在本文中将信任关系与协同过滤推荐相结合,通过用户之间的信任关系扩充用户最近邻集合,从而缓解数据稀疏性问题。③使用Movie Lens数据集来对改进的算法进行实验,然后将实验的结果与现有的一些协同过滤算法进行比较,以此来对算法改进的有效性进行验证。④设计并实现一个电影推荐系统。为了尽可能多地覆盖到用户的兴趣点,系统中实现了基于内容推荐、基于项目协同过滤推荐以及结合信任模型的协同过滤推荐这三种推荐算法。并且为了满足大数据的处理需求,推荐引擎是基于Hadoop分布式平台实现的。更多还原

【Abstract】 Today, the Internet is flooded with all kinds of information, which leads to the increasingly serious problem of information overload. For ordinary users, it’s easy to be disturbed by the complex information. So they have to spend more time and energy to look for the information they need. An effective method to solve the problem of information overload is the recommendation system. Recommendation system can provide personalized recommendation services for users according to the characteristics of them. Recommendation system needs to achieve different recommendation algorithms according different requirements, and the collaborative filtering recommendation algorithm is the most popular and practical recommendation technology among all the recommendation algorithms. In this paper, we will study and improve the existing trust based collaborative filtering recommendation algorithms, and then design and implement a movie recommendation system with multiple algorithms. The main work of this paper includes:① In-depth study on recommendation system, analyze its concept and structure, evaluation standards, etc…Focus on the analysis of process, characteristics of several commonly used recommendation algorithm.② The traditional collaborative filtering recommendation algorithm has the data sparse problem which affects the accuracy of recommendation results in a great extent. In this paper, we combine the user trust model with collaborative filtering. The nearest neighbor set can be expanded by the trust relationship between different users. The data sparse problem can be alleviated by this method.③ Evaluate the improved algorithm with the Movie Lens dataset, and then compare the experimental result with some existing and correlative collaborative filtering algorithms to verify the effectiveness of the improvements.④ Design and implement a movie recommendation system. In order to cover the interest of users as much as possible, the recommendation system implements three algorithms including content based recommendation algorithm, item based collaborative filtering recommendation algorithm and the collaborative filtering recommendation algorithm combined with trust model. And the recommendation engine is realized based on the Hadoop distributed platform.更多还原