电子商务个性化推荐系统的设计与实现-软件工程专业论文.docxVIP

摘 要 在网上购物日益普及的今天，个人相对稳定的购买需求与网站提供的纷繁庞大的产 品之间的矛盾越来越突出，“信息过载”和“信息迷失”的现象日益严重。个性化推荐 技术在电子商务网站上的应用越来越普及，但是随着人们购物要求的提高，其面对的挑 战也日益严峻。 第一：冷启动问题，因为新用户几乎没有可以利用的行为信息，所以很难给出精确 的推荐；另一方面新产品由于被选择次数少，也难以找到合适的方法推荐。 第二：用户行为模式的挖掘和利用，用户的购买行为会随着时间、地点等各种外界 因素而变化，如何准确的把握和利用用户的行为模式信息是推荐系统成败的关键点。 第三：个性化推荐算法的选择，个性化推荐算法主要包括基于内容的推荐和协同过 滤推荐，如何根据网站的数据选择合适的推荐算法直接决定着推荐结果的优劣。 综上所述，本文采用热销榜和标签系统来解决冷启动问题，对于通过搜索引擎到达 产品页面的用户，根据 solr 站内全文搜索系统向用户推荐和其搜索关键词相关的产品； 用户行为偏好的挖掘采取混合挖掘方式，隐式和显式信息共同来决定用户的喜好，并采 用向量空间方法来构建用户行为模型；对于产品特征关键词权重的计算利用 HTTPCWS 中 文分词系统来进行分词处理；针对各个推荐算法的优缺点并依据商城的实际数据提出利 用内容和协同过滤这两种算法组合加权来对用户进行针对性的推荐；最后利用百姓商城 电子商务平台构建一个个性化产品推荐系统。 关键词：电子商务，个性化推荐系统，冷启动，用户行为模式，组合加权推荐算法 I  III IV ABSTRACT With stable growing in the popularity of online shopping, the contradiction between personal purchase behaving and the huge commodity demand that the website provides is more and more prominent. The phenomenon of information overload and information lost is growing much more serious. The application of personalized recommendation technology on e-commerce website is becoming increasingly popular, but with the improvement of peoples shopping requirements, the challenge it faces is becoming more and more rigorous. First: cold start problem, because new users can take advantage of little behavioral information, it is difficult to give precise recommendation; On the other hand the new products have a small number of chance to be selected, it is difficult to find a suitable method to give right recommendation. Second: the mining and utilization of user behavior pattern, digging user behavior patterns can improve the effect of the recommendation to some extent, but the users buying behavior will change with the time, and how to accurately grasp the behavior patterns and making full use of information is a starting point for the success of the recommendation system. Third: The selection of personalized recommendation algorithm, personalized recommendation algorithm including content-based recommendation and collaborativ