A Thesis Submitted in Partial Fulfillment of the Requirements
for the Degree for the Master of Engineering
Design and Implementation of a Movie
Recommender System
搜索附近人Candidate :You FangYuan
保持安静英语Major : Software Engineering
Supervisor : Prof. Shen Gang
刺猬歌词HuazhongUniversity of Science &Technology
Wuhan 430074, P.R.China
January, 2013
蔡佩君
独创性声明
肛门痛什么原因本人声明所呈交的学位论文是我个人在导师指导下进行的研究工作及取得的研究成果。尽我所知，除文中已经标明引用的内容外，本论文不包含任何其他个人或集体已经发表或撰写过的研究成果。对本文的研究做出贡献的个人和集体，均已在文中以明确方式标明。本人完全意识到本声明的法律结果由本人承担。
学位论文作者签名：
日期：年月日
学位论文版权使用授权书
本学位论文作者完全了解学校有关保留、使用学位论文的规定，即：学校有权保留并向国家有关部门或机构送交论文的复印件和电子版，允许论文被查阅和借阅。本人授权华中科技大学可以将本学位论文的全部或部分内容编入有关数据库进行检索，可以采用影印、缩印或扫描等复制手段保存和汇编本学位论文。
保密□，在年解密后适用本授权书。
本论文属于
手工礼物送女生
不保密□。
（请在以上方框内打“√”）
学位论文作者签名：指导教师签名：
日期：年月日日期：年月日
摘要
随着互联网与移动互联网迅速普及，网络上的电影娱乐信息数量相当庞大，人们对获取感兴趣的电影娱乐信息的需求越来越大，个性化的电影推荐系统成为一个热门。然而电影信息的表示相当复杂，已有的相似度计算方法与推荐算法都各有优势，导致单一的相似度计算方法与推荐算法无法合适地应用于电影推荐系统中。大量的电影数据的管理运营随着数据量的增长也变得越来越复杂，因此，如何综合各种算法的优势给用户可靠的电影推荐结果，并保证用户能访问到正确的推荐数据成为推荐系统设计中需要解决的一个重要问题。
系统在推荐模块中使用了更切合实际的电影相似度计算方法，结合协同过滤算法与基于内容的过滤算法，有效的解决了系统的冷启动问题和推荐的准确性问题，推荐算法使用集中平均法来预测用户对电影的评分，避免了用户个人评分习惯对预测评分产生的不利影响。将数据集，推荐引擎，评分预测器，相似度计算器等重要部件高度分离，并提供各种有效的具体实现算法派生类。在运营模块中将系
统分为测试环境与正式环境两部分，在测试环境中进行测试，然后同步到正式环境中，使用新型同步算法与错误检测算法提高系统的准确性与效率。
系统推荐模块的可扩展性非常强，根据具体的数据集可选择适合的评分预测器与相似度计算器，可以使用系统自带的经典算法，也可以自定义推荐算法，使得推荐结果的准确性大大提高。在运营模块中使用数据库依赖同步算法与错误检测算法使得数据的运营准确性与效率大大提高。
我亲爱的祖国关键词：推荐系统协同过滤基于邻域推荐平均集中法
Abstract
With the rapidly growing popularity of the Internet and mobile Internet, the number of movie entertainment information on the network has been extremely substantial, more and more people get interested in movies and entertainment information, personalized movie recommendation system has become a hot. However, movie information reprentation is quite complex, the existing similarity calculation method recommended algorithms have their own advantages, resulting in a single similarity calculation method recommended algorithm can not be properly applied to movie recommendation system. Management and operation of a large number of movie data has become increasingly complex with the growth of the amount of data, therefore, how to integrate the advantag
es of various algorithms to generate reliably movie recommendation results and ensure that urs can access to the correct recommended data become important issues need to be addresd in the recommended system design.
System module is recommended to u the more realistic movie similarity calculation method, combined with collaborative filtering algorithm and content-bad filtering algorithms to effectively solve the problem of the accuracy problem and cold start problems of the system, recommended algorithm us mean-centering method to predict ur ratings of movies, to avoid the adver effects of the ur's personal rating habits prediction score. Important component such as datat, recommendation engine, rated predictor similarity calculator are of a high degree of paration, with providing a variety of effective implementation algorithm derived class. In the operator module in the system is divided into two parts, the test environment and formal environment, the operator test the system well in the test environment and then synchronize to a formal environment, using the new synchronization algorithm with error detection algorithm to improve the accuracy and efficiency of the system.
The scalability of the recommendation module is very strong, so you can lect suitable score predicted similarity calculator depending on the data t, you can u the classic algorithm which co
mes with the system or you can also define custom recommendation algorithm, this can makes recommendation result accuracy greatly
incread. In the operator system module with the synchronization algorithm with error detection algorithm operators greatly improve the accuracy and efficiency of data using the databa module operators.
Key words：recommender system collaborative filtering
思想品德教育Neighborhood-bad recommendation mean-centering method
