Title: A user-concept matrix clustering algorithm for efficient next page prediction

Authors: Wedad Hussein; Tarek F. Gharib; Rasha M. Ismail; Mostafa Gadal-Haqq M. Mostafa

Addresses: Faculty of Computer and Information Sciences, Ain Shams University, 11566, Cairo, Egypt ' Faculty of Computer and Information Sciences, Ain Shams University, 11566, Cairo, Egypt; Faculty of Computing and Information Technology, King Abdulaziz University, Jeddah 21589, Saudi Arabia ' Faculty of Computer and Information Sciences, Ain Shams University, 11566, Cairo, Egypt ' Faculty of Computer and Information Sciences, Ain Shams University, 11566, Cairo, Egypt

Abstract: Web personalisation is the process of customising a website's content to users' specific needs. Next page prediction is one of the basic techniques needed for personalisation. In this paper, we present a framework for next page prediction that uses user-concept matrix clustering to integrate semantic information into web usage mining process for the purpose of improving prediction quality. We use clustering to group users based on common interests expressed as concept vectors and search only the set of frequent patterns matched to a user's cluster to make a prediction. The proposed framework was tested over two different datasets and compared to usage mining techniques that search the whole set of frequent patterns. The results showed a 33% and 2.1% improvement in the average system accuracy as well as 6.6% and 7.3% improvement in the average system precision and a 6.5% and 1.7% in coverage for the two datasets respectively, within the same computation timeframe.

Keywords: recommender systems; web usage mining; semantic web mining; user-concept matrix; clustering algorithms; next page prediction; web personalisation; recommendation systems; semantics; pattern matching; data mining.

DOI: 10.1504/IJKWI.2016.078718

International Journal of Knowledge and Web Intelligence, 2016 Vol.5 No.3, pp.208 - 229

Received: 17 Sep 2015
Accepted: 27 Jan 2016
Published online: 01 Sep 2016 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article