Personalising e-commerce: An approach based on capturing real-time browsing behaviour

The rise in online trading and consequent proliferation of user’s interests on the internet has created significant interest in the notion of personalised product recommendations. This research describes a Web personalisation and recommender system that uses web usage data mining techniques to predict user’s preferences and interests during each browsing session and provide tailored content based on this prediction in order to leverage customer retention and engage users at personal level. There are problems associated with traditional Web personalisation approaches such as collaborative or content-based filtering. These problems include lack of scalability and reliance on subjective user input during online registration process.

In this research, we present a framework distinguishing the web usage mining task with a short-term (online) module and a long-term (offline) module. An offline web usage mining operation is scheduled at off-peak times acting upon click stream, whilst, an online mining process is dedicated to delivering tailored personalised services to the current user in real time.

This research demonstrates that such an approach can overcome the scalability problem that is common among traditional approaches, and successfully provide a personalised service to any anonymous visits without a dependency on user registration. In addition, the proposed system is implemented in an open and expandable platform based on a three-tier architecture. This research has demonstrated that the three-tier implementation can overcome the limitations of traditional systems based on a two-tier architecture with increased scalability.