【作者】 孙斌；

【导师】 薛志东；

【作者基本信息】 华中科技大学 ， 软件工程， 2012， 硕士

【摘要】 互联网已经逐渐走入人们的生活，人们对信息的利用也逐渐变的全球化，通过网络来获取信息变的日常化。近年来，互联网信息数据量爆炸式增长，海量信息充斥着互联网，用户越来越难准确找到自己所需要的信息，―信息过载‖问题成为互联网发展的巨大挑战。个性化推荐技术作为目前解决信息过载最有效的工具之一。通过推荐引擎能够帮助用户智能过滤信息。基于目前行业比较流行的推荐技术，协同过滤推荐技术、内容过滤推荐技术能够进行一定程度的智能推荐。但是在实际应用过程中依然面临的诸如推荐质量低、自动化程度差、冷启动问题、实时响应差等挑战。通过分析单一协同过滤和单一内容过滤技术的优缺点，在基于协同过滤的技术中，融合内容过滤技术，提出了混合推荐算法的思想。利用项目内容弥补用户对项目的缺失评分，从而使用内容过滤为用户寻找相似邻居，然后就可以使用协同过滤技术产生推荐结果。从理论上通过使用KNN算法的内容过滤技术对稀疏矩阵进行补充，能够提高协同过滤推荐的质量，从而提高最终推荐质量。基于以上混合推荐算法的思想和理论，设计异步在线、离线模块，实现了混合推荐算法，并最终通过平均绝对偏差的衡量标准进行结果评估，验证了混合推荐算法优越性。最后，设计实现了一个基于混合推荐算法的电影推荐系统。该系统基于探讨的混合推荐算法，提供个性化推荐服务，并综合了用户行为记录的方法，从不同角度收集用户的兴趣爱好，获取用户信息，提升了系统的用户体验。更多还原

【Abstract】 The Internet has been gradually going into people’s lives, the globalization ofinformation has become a trend, through the network to obtain information becomecommon.In recent years, the Internet information grow quickly, vast amounts ofinformation filled with the Internet, users feel more and more difficult to accurately findthe information they need,"information overload" has become a huge challenge for thedevelopment of the Internet.Personalized recommendation technology is one of the most effective tools to solvethe information overload.As recommendation engine can help users filter informationintelligently.Collaborative filtering to recommend the technology and content filteringrecommendation is popular recommendation technology, they can filter information forthe user.But in the actual application process is still facing problems such as the lowquality of recommended, degree of automation, cold start, real-time response.By analyzingthe advantages and disadvantages of a single collaborative filtering and a single contentfiltering technology, based on collaborative filtering technology, integration of contentfiltering technology, put forward the idea of a hybrid recommendation algorithm.Usingproject’s content to compensate for the lack of ratings of users on the project, which usethe content filtering for the user to find similar neighbors, then you can use collaborativefiltering technology to produce the recommended results.KNN algorithm-based contentfiltering technology to supplement the sparse matrix can improve the quality ofcollaborative filtering recommendation, thereby enhancing the final recommendationquality.Based on the above hybrid recommendation algorithm ideas and theory and designonline, offline module,we can make mixed recommendation algorithm, by the measure ofthe average absolute deviation of the results of assessment, it verify the superiority of thehybrid recommendation algorithm.Finally, design a movie recommendation system based on a hybrid recommendationalgorithm. The system is based on the hybrid recommendation algorithm explored,provide personalized recommendation services, and comprehensive record of userbehavior, from different perspectives to collect users’ interests and hobbies, to get user information to enhance the user experience.更多还原