基于聚类算法的购物篮压缩研究

Research on Shopping Basket Compression Based on Clustering Algorithm

购物篮分析是数据挖掘技术在零售业的典型应用之一,旨在从零售交易记录中分析出顾客经常同时购买的商品组合,挖掘出购物篮中有价值的信息。然而实际分析中往往得到的是数以千计的购物篮,企业很难从这数量众多的购物篮中找到真正感兴趣和有价值的,这给实际的应用造成了很大障碍。针对传统挖掘方法得到购物篮数量过多的问题,定义了一系列特征属性表示购物篮,提出了一种基于K-Means层次聚类算法根据属性值对购物篮进行压缩的方法。该方法通过对真实购物篮进行实验研究与分析。为验证提出方法的有效性和可行性,将其与传统压缩方法进行了对比。实验结果表明,相对于其他传统压缩方法,由提出的压缩方法筛选得到的购物篮具有更高的有效性和实用价值,并达到了压缩购物篮集合的效果。

The analysis of shopping basket is one of the typical applications for data mining technology in the retail industry,which aims to analyze the combination of goods which customers frequently buy from the retail transaction records and dig out the valuable informa- tion in the shopping basket. However,the thousands of baskets often exist in actual analysis. It is difficult for enterprises from this large number of shopping basket to find that with real interest and value, which brings a great obstacle to application. In view of problem of ex- cessive baskets in traditional mining methods,a series of characteristic attribute are defined for representation of shopping baskets,and a method of K -Means-based hierarchical clustering algorithm compressing the shopping basket according to the attribute value is presen- ted. In order to verify its effectiveness and feasibility, the comparison is made between the proposed and traditional method. The experi- ment shows that after comparison,the shopping baskets from proposed method own the higher availability and application value with the effect of compressing the shopping basket sets.