Personalization and recommendation systems play an important role in online businesses. In order to provide personalized services effectively, a firm needs to decide how users should be modeled, determine the values of important profile characteristics for each user, and then recommend products or services that are best matched to the user at the appropriate time. My dissertation examines several aspects of the personalization process.;In the first essay, I show how a firm can learn relevant customer profiles from interactions with the customers. I focus on customer profiles that consist of market segmentation constructs such as demographic and psychographic characteristics, as they are especially useful in deploying target marketing strategies. I develop a probabilistic framework for learning such profiles from the navigational history of customers. I show that the performance of the proposed approach is as good as, and often better than, other established classification techniques. The second essay examines how a site could accelerate learning a customer's profile, so that it could benefit from offering personalized services sooner. The ability to learn a customer's profile depends to a large extent on the pages visited (or links clicked) by the user. I develop a model to determine the offer set for each page visited by the user that can help accelerate the learning process. I show that the proposed model vastly accelerates the learning of the user profiles.;The third essay examines how a firm can maximize its expected payoffs when making recommendations to users by leveraging the knowledge of the profiles of visitors to the site. I develop a methodology that accounts for the interactions among items in an offer set in order to determine the expected payoff. Identifying the optimal offer set is a difficult problem when the number of candidate items is large. I develop an efficient heuristic for this problem, and show that it performs well for both small and large problem instances.
展开▼
