【作者】 张颖；

【导师】 丁宇新；

【作者基本信息】 哈尔滨工业大学 ， 计算机技术（专业学位）， 2015， 硕士

【摘要】 随着电脑的普及和互联网的快速发展,人们获取信息的渠道越来越多,最终导致人们获取到的信息由最初的过分简单和单一,变成现如今的信息过分富足。面对庞大的信息量,人们想要精准快速的找到所需变得特别困难。为了解决这一问题,首先出现的是搜索技术。搜索技术带来的新问题是无法提供可定制化的服务,反馈结果的精确性也有待提高,在这种背景下,推荐技术出现了,与搜索技术相比,推荐技术最显著的特征就是可以提供定制化服务。目前,推荐技术被应用于各个领域,比如商品推荐、广告推荐、新闻推荐等。传统的推荐系统一般采用两种推荐算法:协同过滤推荐算法和基于内容的推荐算法。这两种算法均存在不足,协同过滤推荐算法存在的两个明显问题:冷启动问题和随着矩阵增大带来的评分矩阵稀疏性问题;基于内容的推荐算法需要进行大量的文本计算,为了规避两种算法的缺点,最大限度发挥两种算法的优势,本文提出了混合推荐算法思想,目的在于充分利用两种推荐算法的优势,为用户提供更个性化和可定制的推荐服务。本文的工作从以下几个方面展开:首先对本领域的相关研究工作进行了较全面的总结与分析,重点研究基于内容推荐的相关计算算法和基于协同过滤推荐的相关计算算法,以及利用这些算法如何实现推荐系统。研究新闻推荐算法中新闻主题词提取的相关理论和算法。针对协同过滤算法存在的矩阵稀疏性和冷启动问题,研究如何将多种过滤算法混合形成高效的混合推荐算法。利用混合的相关知识,设计和实现了本文的新闻推荐系统,并对系统算法的有效性进行评测。更多还原

【Abstract】 With the popularity of computers and the rapid development of Internet, people can collect information from many different ways, which results the information overloading problem. Facing a large amount of information, it is very difficult for people to quickly and accurately retrieve the information they want. To solve this problem, traditional searching technique was proposed. However, the problems of the traditional searching technique are that it cannot provide customized service and has not high retrieval accuracy. Under this situation, recommendation algorithm was proposed. Compared with information retrieval, recommendation algorithm can supply customized service. In present recommendation algorithm has been widely applied in different fields, such as goods recommendation, news recommendation and advertisement recommendation.Traditional recommendation systems use two algorithms: collaborative filtering recommendation algorithm and content-based recommendation algorithm. However, these two algorithms have their own drawbacks. The collaborative filtering recommendation has the cold start problem and the scoring matrix sparsity problem with the increase of users. The recommendation algorithm based on the contents of the text has a high calculation complexity. In order to circumvent the drawbacks of the two recommendation algorithms and maximize the advantages of the two algorithms, this paper proposes a hybrid recommendation algorithm. This method is aimed at fully utilizing the advantages of the two recommendation algorithm s to provide users personalized and customized services. The main works of this paper are as follows:Firstly, we introduce and analysis the research works in this filed. We focus on the content-based recommendation algorithm, the collaborative filtering recommendation algorithm, and how to implement a system using these algorithms.Study the algorithm for extracting topic words of news.To solve the matrix sparse problem and the cold start problem existed in collaborative filtering recommendation algorithm, we study how to combine the content-based filtering algorithm and collaborative filtering recommendation algorithm to form an efficient hybrid recommendation algorithm.We design and implement a news recommendation system based the hybrid recommendation algorithm, and evaluate the effectiveness of the algorithm.更多还原