Data Preprocessing Techniques in Web Usage Mining: A Literature Review

Abstract

With the massive growth of Web, there is a huge volume of dynamic, distributed, heterogeneous, structured, unstructured, semi-structured, and high dimensional data available on Web. Apart from content and structural information of Website, server logs are also considered as a valuable source of information. Web usage mining is a class of Web mining where users’ navigation behavior is analyzed from Web server log. It is divided into three phases: Data preprocessing, pattern discovery, and pattern evaluation. Among them, data preprocessing is considered as a time-consuming and complex phase in Web usage mining process due to huge and noisy nature of log data. This article present a review and critical analysis of sequential techniques applied in data preprocessing of Web server log with emphasis on sub-phases such as data cleaning, user identification, and session identification. Moreover this article also includes the survey of techniques applied for server log analytics using Big Data technologies such as Hadoop MapReduce and Spark framework. This article would be helpful for researchers to find issues related to data cleaning, user identification, and session identification phases of Web usage mining process.

Keywords: Data Preprocessing, Web Usage Mining, Data Cleaning, User Identification, Session Identification, Big Data, MapReduce