The Phases of Web Usage Mining for Classification and Pattern Analysis

Abstract

The Internet is an enormous store of web pages and links between them. It gives a tremendous measure of data to internet customers. The advancement of the web is extraordinary as around one million pages are included each day.  Users access information such as the weblog records are expanding rapidly and the storage also in terms of terra bytes. Web Usage Mining (WUM) relates to extracting the information from these log files to analyse the performance of users at numerous applications such as e-commerce systems.  The conventional WUM techniques contain mainly three states such as pre-processing, detecting the patterns, and learning or analyse these patterns. Among which, pre-processing and pattern analysis is crucial for many applications since the log data is commonly noisy and inconsistent. Many data mining techniques exist to extract the frequent patterns, associations, and correlations from these log data, still, WUM poses many challenges such as scalability, accuracy, and privacy.  This paper presents a general review on WUM and its significance for the creators and those intrigued by online stores and personalization of their websites. Then, the recent works on WUM to discover the interesting patterns from the weblog data and web recommendation systems. Furthermore, the steps involved in the WUM along with the issues and future directions also presented.