个性化推荐搜索引擎的设计与实现

发布时间：2018-04-21 19:07

本文选题：个性化推荐 + 用户标注　；参考：《电子科技大学》2012年硕士论文

【摘要】：随着现代电子商务和网络的快速发展，各大商务网站都为用户提供了越来越人性化个性化的服务，个性化的推荐搜索引擎的研究也越来越广泛的被应用。目前各大商务网站的个性化推荐引擎基本都是统一推荐或者是针对热门类目进行推荐，没有设计出一种针对不同的用户给出不同的推荐信息的推荐搜索引擎，这种搜索引擎就像是了解你性格兴趣的商品推销员，能够总是在你需要某些商品信息的时候第一时间把推荐展示在你面前。本文针对上述问题，在分析和研究主流的搜索引擎的基础上设计出一个简易的支持短文本的个性化推荐搜索引擎系统包括OnLine模块和OffLine模块。本系统能够根据不同的注册用户，并针对该用户的兴趣爱好给出个性化的推荐信息。本系统研究的主要内容为： 1.在线OnLine处理模块，其中包括Http服务器，query分析器和用户特征获取器，Rank评分核心机制，排序评分排序等几个部分。设计一个简单Http服务器来作为本系统的一个服务器容器，由于本文研究设计的个性化推荐搜索引擎系统是一个轻量级的系统，因此需要一个同样简易化轻量级的Http网络服务器来支持。通过用户特征获取器来获取该用户的基本信息和兴趣爱好，query分析器用来获取用户查询记录中相关记录处理后的一个倒排表。Rank评分核心机制，也是本系统的核心，对查询分析和用户特征获取器获取的所有数据记录进行评分处理，依据分数排序，获取一定数量排在前面的结果集，即是根据用户的兴趣爱好和购买历史所推荐的优先结果集，优先返回和用户兴趣爱好相关性较大的记录。 2.离线OffLine模块，该模块主要负责后台数据的处理。后台需要处理的数据主要是用户标注模块和查询标注模块。用户信息分为基本信息和行为信息，用户的基本信息通过实验数据的方式获得，基本信息中包含用户的最基本的特征，，而用户的行为信息中一般分为兴趣和购买历史两个部分，这些信息反映了用户的兴趣爱好等特征，通过对用户的信息分析和用户查询记录的分析，过滤对本次查询无效的记录。
[Abstract]:With the rapid development of modern electronic commerce and network, each major business website provides more and more personalized service for users, and the research of personalized recommendation search engine is more and more widely used. At present, the personalized recommendation engine of each major business website is basically a unified recommendation or a recommendation for hot categories. There is no design of a recommendation search engine for different users to give different recommendation information. This search engine is like a merchandiser who understands your personality interests and can always present recommendations to you the first time you need information about something. In this paper, based on the analysis and research of the mainstream search engine, a simple personalized recommendation search engine system supporting short text is designed, which includes OnLine module and OffLine module. The system can give personalized recommendation information according to different registered users and their interests. The main contents of this system are as follows: 1. Online OnLine processing module, which includes Http server query analyzer and user feature acquirer Rank scoring core mechanism, ranking rating ranking and other parts. A simple Http server is designed as a server container of this system. Because the personalized recommendation search engine system studied in this paper is a lightweight system, Hence the need for an equally easy lightweight Http web server to support. The basic information and interests of the user are obtained by the user feature acquirer. The query analyzer is used to obtain an inverted table. Rank scoring core mechanism after the processing of the related records in the user query record, which is also the core of the system. All the data records obtained by the query analysis and the user feature acquirer are graded, and according to the ranking of the scores, a certain number of the first result sets are obtained, that is, the priority result set recommended according to the user's interest and purchase history. Priority return and user interests related to the record. 2. Offline OffLine module, this module is mainly responsible for background data processing. The data needed to be processed in the background are the user tagging module and the query annotation module. User information is divided into basic information and behavior information, the basic information of users is obtained through experimental data, the basic information contains the most basic characteristics of users, and the behavior information of users is generally divided into two parts: interest and purchase history. This information reflects the characteristics of the user's interests and hobbies. Through the analysis of the user's information and the analysis of the user's query records, the invalid records of this query are filtered.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP391.3

【参考文献】