Product Recommendation for E-Commerce Application via Web Server Logs

Abstract

E-commerce applications use product recommendation on their application for increase sales. Application owners use their own business experience and business related assumptions for identify recommendations. These recommendations not accurate most of time and it leads application consumers with multiple choices for a specific product which leads to product overload. These incorrect recommendations badly affected for archived expected sales targets and give bad user experience. Therefor it is important to identify most accurate recommendations and present them on store front. This developed system focused on web usage mining. System extract product association rules generated from machine learning tool and integrate those findings with E-Commerce application without user involvement. System use weka-3-6-12 as machine learning tool. Scope of research includes access log data acquisition, data pre-processing, data cleaning, product based association rule generation and integrate identified recommendations with E-Commerce application. Selected E-commerce application for integrated with developed system build on Magento e-commerce platform (Enterprise version 12) and hosted on Ubuntu server with Apache 2.0, PHP 5.5 and MySQL. System generated recommendations evaluate with already used recommendations using data analytic tools with dependent sample t-Test. Test results prove system generated recommendations are more accurate and relevant compare with already used recommendations.