Nowadays web sites have become an important means for communication. Every application is converted to internet based application. A huge amount of the data is added daily and huge amount of information is accessed from the web. The millions to billions of web users are accessing the interested data from the web making the network traffic. To overcome these problems, recommender systems which are the part of machine learning are introduced. At present, recommender systems play an important role in the e-commerce field. The shopping sites where the users are recommended with the interested products and resources based on their navigation behavior, profile and their interest on the products. The recommender systems are basically divided into three types, they are content based, collaborative and hybrid recommender systems. As the web users are increasing the network performance is affected. Thus it is required to enhance the performance of the network by using the mining and machine learning classification techniques. In this paper, we propose an efficient web recommender system, which uses the concept of preprocessing, clustering technique and expectation maximization naïve bayes as predictive model.

Web access logs Preprocessing K-Means clustering Expectation Maximization Naïve Bayes Classifier.