Recommendation System Based on Optimal Feature Selection Algorithm for Predictive Analysis

Abstract

The Internet business model with predictive analysis provides efficient way of accessibility to customers with less resources and minimum expenses. User access log files are considered in this work for predictive analysis. But the log files contain lots of unwanted information and not in a form for predictive analysis. So, the web logs need to be pre-processed and further handled by the recommendation system for predictive analysis. In the processing phase, unwanted details are removed; generated user sequences, session sequences and complete user access path patterns are also generated. In the proposed recommender system, significant feature set is considered using our proposed feature selection algorithm to reduce the computational cost and to improve the quality in the predictive analysis process. In this paper, vizhamurasu news web server user access logs are considered for predictive analysis. Our proposed feature selection algorithm performance is analyzed using the performance metrics. The performance analysis shows that our proposed algorithms significantly improve the prediction accuracy than the existing.