Discovering Rules for Predicting Customers’ Attitude toward Internet Retailers

Customers’ attitude toward online shopping is the key to the survival and profitability of Internet retailers in the intensely competitive market. Aiming to automatically discover knowledge for predicting customers’ attitude toward Internet retailers in relative to traditional retailers, we examine two classification approaches, decision trees and neuro-fuzzy systems, which are capable of generating such knowledge in form of rules. The neuro-fuzzy approach has rarely been investigated in the Internet retailer context. We compare the two knowledge discovery approaches using data sets for two types of products collected from an empirical survey. The results show that while the performance of the two approaches is comparable, the neuro-fuzzy model is superior to the decision tree in handling uncertainty and imprecision of the data sets. Motivated by the potential value that knowledge discovery and Web services can add to existing services, we also propose an architecture that enables agile Web services in Internet retailing. The insights on methodology and the Internet retailing application gained from this study suggest a number of interesting issues for future research.