Improving Link Recommendation through Users Clustering

Link recommendation is a business activity that is critical in attracting customers. Accordingly,improving the quality of a recommendation to fulfil customers’ needs is important in fiercelycompetitive environments. Studying users’ behaviours in the past with web usage mining techniquesutilization can be worthy help in link recommendation affair. In this paper, users’ behaviouralpatterns are obtained from studying of users’ access log by applying web mining techniques. One ofthe innovative aspects of the research is selecting some behavioural features from users which arestored in users’ profiles. Users are clustered according these features with K-means algorithm thenuseful knowledge can be extracted from web user access patterns. Neural network usage is anotherfeature of proposed system to form recommender engine, which its function is to find properbehavioural pattern for users’ session and forecast the links that users are likely want to see it. Asresearch conclusion presents recommender engine has the appropriate accuracy in prediction oflinks that user are most likely want to see them.

Keywords:

Web mining , Web usage mining , recommender engine , Clustering , Neural network