An artificial neural network enhanced Semantic Variable Length Markov Chain Model (ANNSVLMC) for web navigation session mining

Markov models are widely utilized for the purpose of analyzing user web navigation sessions. A novel SVLMC (Semantic Variable Length Markov Chain Model)was developed in which recommendations were generated by incorporating semantic data of web pages with navigation history. This technique duplicate the state by adaptingstate cloning concept in a way that separates in-links whose corresponding second-order probabilities diverge, and incorporates a clustering technique to assignin-links with similar second-order probabilities to the same clone. However, SVLMC resulted in high state space complexity while applying cloning approaches in higher order Markov model. ANNSVLMC(Artificial neural network based Semantic Variable Length Markov Chain Model) is developed by incorporating ANN model into the SVLMC. The forecast state predicted by ANN is integrated into the transition matrix of the Markov chain where estimated forecast accuracy of the ANN model is used as the transition probability from the current state to the forecast state. Thus the combined Markov chain process with the ANN models provides valuable information about users interest on next page from the web usage session. The experimental result shows better performance of ANNSVLMC than SVLMC.