Efficient mining techniques for web users behaviour

Abstract

Web usage mining has emerged as an extension to web mining. This computing technology is mainly used to discover the knowledge about the web users interest. Mining of web users information is leveraged by various resources. The resources may be web log file, web ranking data, web rating data and web review data. Today, web usage mining is performed using a variety of technologies and tools. This is helpful in providing the necessary information or product to the web users. This thesis mainly focuses on discovering knowledge more accurately by effectively utilizing the data. The heterogenous data, along with the hybrid method is used to discover the knowledge accurately and efficiently. The web log data, web ranking data, web rating data and web review data are used to discover knowledge. Three different hybrid approaches are proposed to find an effective method in discovering knowledge about the web users. The different methods are hybrid aggregation using weighted approach, adaptive polling-based ensemble method to discover knowledge and genetic kernel difference based fuzzy c-mean clustering method. The thesis initially commences with the discovery of knowledge through homogeneous methods such as web log, web ranking, web rating and web review. Then the heterogeneous method is implemented by combining the homogeneous methods. The hybrid aggregation using weighted approach is used to combine the knowledge discovered through various homogeneous approaches. The aggregation is performed based on moving average based approach. Then the adaptive polling-based ensemble method is used to discover knowledge. This method is an optimization-based method where this method removes the irrelevant methods and assigns weight accordingly to the most relevant methods. newline newline newline