Extraction of interesting rules from internet search histories

Rule extraction aims to ultimately improve business performance through an understanding of past and present search histories of customers. A challenging task is to determine interesting rules from their heterogeneous search histories of shopping in the Internet. For this purpose neural network (NN) and canonical correlation analysis (CCA) are used. Customers visit Web pages one after another and leave their valuable search information behind. Firstly we produce a homogeneous data set from their heterogeneous search histories. It is difficult task to produce a homogeneous data from heterogeneous data without changing their characteristics of data. Secondly these data are trained by unsupervised NN to get their significant class. Thirdly we extract the maximally correlated customers by using CCA and then interesting rules are extracted among their maximally correlated customer. This is important for the traders, marketers and customers for making future business plan.