An Enhanced Web Page Recommendation System Using Hidden Markov Model and Page Rank Technique

Abstract

Abstract

The rapid expansion of the World Wide Web (WWW) has created an opportunity to disseminate and gather information online. There is an increasing need to study the behaviour of web users to serve them better by reducing the access latency using efficient web prediction technique. Markov Models have been widely used for predicting next web page request from the users’ navigational behaviour recorded in the web log. This usage-based technique can be combined with the structural properties of the web pages to achieve better prediction accuracy. This study combines both Markov Model and Page ranking, which considers the structural properties of the Web. In order to create an efficient prediction model, the original data was preprocessed in the form that can be used for unsupervised learning. The pre-processed data was then analyzed using unsupervised learning K-means clustering algorithm. To increase the efficiency of Hidden Markov Model (HMM), efficient ranking algorithm was used to identify the most relevant page in clusters. i.e. PageRank. The HMM was then used to predict users web navigation path. Briefly, results from the study shows that Hidden Markov model and Clustering can work together and provide better

prediction results without compromise to accuracy though with a trade-off in time complexity, HMM is more accurate for predicting navigational paths thereby enhancing web page recommendation.

Download   View