Novel Methods for Prediction of Customer Purchase Probability

Web is a huge repository for storing different types of information. As the number of webpage's and websites has increased rapidly, understanding and discovering web user’s behavior are essential for the web monitoring and recommendation systems. user’s behavior in the web can be predicted using the online databases such as log files and prediction models. The main problems of user’s behavior analysis over online databases are: lot of transactions between user and the web and the volume of data which are not completely structured. knowing the customer behavior for online product recommendation and prediction of customers purchase probability using click stream data and tracking of customer behavior is essential part of online recommendation system. This paper presents an efficient approaches that works using Bayesian belief networks (BBN) and Nearest-neighbor collaborative filtering that calculate the probabilities of inter-dependent events and provides a successful means of generating recommendations for web users.