电子商务协同推荐算法的设计与应用

The Design and Application of E-commerce Collaborative Recommendation Algorithm

电子商务的快速增长导致产品过多,而网络上的客户对他们所接触的产品很难取舍。为了解决此问题,各种推荐方法应运而生。协作过滤(CF)是其中最成功的推荐方法,其被广泛应用于电子商务中,然而这种方法的稀疏性和可扩展性可能导致推荐的结果较差。提出了一种基于Web使用挖掘和产品分类的推荐方法,以提高当前基于CF方法的推荐系统的推荐质量和系统性能。Web使用挖掘通过跟踪客户在Web上的购物行为来填充评级数据库,从而产生质量更好的建议。产品分类法用于通过评级数据库的降维来提高搜索最近邻居时的性能。对实际电子商务数据的几项实验表明,与其他CF方法相比,所提出的方法提供了更高质量的建议和更好的性能。

The rapid growth of e-commerce has led to too many products,and customers on the network have difficulty in choosing the products they are exposed to.In order to overcome this problem,various recommended methods have emerged.Collaborative Filtering(CF)is one of the most successful recommended methods,and it is widely used in e-commerce.However,it also exposes some well-known limitations,such as sparsity and scalability,which may result in poor recommendation.This paper proposes a recommendation method based on Web usage mining and product classification to improve the recommendation quality and system performance of the current CF-based recommendation system.Web usage mining populates the rating database by tracking customer shopping behavior on the Web,resulting in better quality recommendations.The product taxonomy is used to improve the performance of searching for nearest neighbors by reducing the dimensionality of the rating database.Several experiments on actual e-commerce data show that the proposed method provides higher quality recommendations and better performance than other CF methods.