基于商品分类的电子商务推荐系统设计

摘要

Virtual property of the network causes difficulties in creating trustable relationship and hence in trade decision.The customised recommendation system can deal with the subjective factors in trust evaluation processing and thus improve the accuracy of trade forecasting. In the paper we establish a fuzzy reputation management system based on the thought of collaborative filtering,which emphasises the processing and customised sharing of the first-hand data.On general node,a two-layer fuzzy reasoning logic is applied for adapting to the human thinking habit,and the trust evaluation is customised by the adjustment on reliability attribute of the nearest neighbour.Through super node we calculate and aggregate the global variables of the reputation value of merchandise categories and the recommendation power,and design the valued fuzzed function according to power-law distribution rule.In end of the paper we illustrate the feasibility of the system application through a given example.%网络的虚拟性导致信任关系难以建立，交易决策困难。个性化推荐系统可以处理信任评估中的主观因素，提高交易预测的准确性。基于协同过滤的思想建立一个模糊信誉管理系统，突出对一手信息的处理与个性化共享。在一般节点上，采用二层的模糊推理逻辑适应人类的思维习惯，并通过对最近邻可靠值属性的调整定制信任评估；通过超级节点计算和汇总商品分类的信誉值、推荐力等全局变量，依据幂律分布规律设计取值的模糊化函数。最后通过示例说明了系统应用的可行性。