Mining Web Log Data for Personalized Recommendation System

Increase in number of internet users in Indonesia boosts the development of e-commerce platform. Whereas potential access to a larger and more diverse customer base is generally viewed as an opportunity, it can also represent increase in competition among e-commerce platforms. Hence, e-commerce needs to develop sophisticated strategies to attract and retain customers, one of which is done through personalization in web services. Recommendation system, one form of web service personalization in e-commerce platform, predicts user preferences and helps them find products that they may be interested in by implementing web mining techniques. This empirical research investigated user web log data which illustrate behavior and implicit preferences of customers in one of e-commerce in Indonesia to predict user preferred product category in their future request. In this study, model-based recommendation system was built based on users' activity in a session and site type using C5.0 algorithm of decision tree technique. Top N recommendations were given based on probability-based ranking of categories resulted from probability estimation of the decision tree.