基于图论和信息最大化保留的在线推荐方法

On-line recommendation method based on graph model and maximizing information retention

随着电子商务的发展,研究一套高效准确的推荐方法不仅便利了网上购物,也有助于加速商品流通,促进经济发展.既有的方法主要从商品的相似性或顾客的相似性出发进行推荐,没能将两者很好的结合,不能充分利用既有的评价信息.鉴于此,提出了基于图论的推荐方法,将人和物的相似性信息结合起来,构成综合的评估图模型,并转化为与之等价的评估矩阵.在评估信息最大化保留的优化目标下,以评估矩阵为基础建立推荐算法,并与既有的推荐方法进行比较.实验结果表明:本文的方法具有计算时间短、准确度高的特点,可以用于实时的在线推荐.

The development of E-commence calls for an effective and accurate recommendation method which not only convinces customers,but also accelerates circulation of commodities and promotes economic development.The existed recommendation methods paid attention to either the similarity of goods or that of customers,thus could not trade off the two aspects of information and make full use of them. In view of the above,this paper proposed recommendation method on the basis of graph model,which synthesized the similarity of customers and goods.The method built a comprehensive assessment model able to be transformed into its equivalent evaluation matrix and established an algorithm based on the above evaluation matrix with the aim of maximizing the retention of information.What＇s more,this paper compared it with the benchmark methods.As a result,the numerical experiments show that the method has short-time calculations,high accuracy and is proper for real-time online recommendation.