Probabilistic user behavior models in online stores for recommender systems

Recommender systems are widely used in online stores because they are expected to improve both user convenience and online store profit. As such, a number of recommendation methods have been proposed in recent years. Functions required for recommender systems vary significantly depending on business models or/and situations. Although an online store can acquire various kinds of information about user behaviors such as purchase history and visiting time of users, this information has not yet been fully used to fulfill the diverse requirements. In this thesis, we propose probabilistic user behavior models for use in online stores to tailor recommender systems to diverse requirements efficiently using various kinds of user behavior information. The probabilistic model-based approach allows us to systematically integrate heterogeneous user behavior information using rules of the probability theory. In particular, we consider three requirements for recommender systems: predictive accuracy, efficiency, and profitability. Models that can accurately predict present user behavior, rather than past user behavior, are necessary for recommendations because behaviors may vary with time. We propose a framework for learning models that best describes present samples and apply the framework to learning choice models that predict the next purchase item. In the proposed framework, models are learned by minimizing a weighted error over time that approximates the expected error at the present time. Efficiency is also an important issue for recommender systems because the systems need frequent updates in order to maintain high accuracy by handling a large number of purchase history data that are accumulated day by day. We present an efficient probabilistic choice model using temporal purchase order information. Fast parameter estimation and high predictive accuracy are achieved by combining multiple simple Markov models based on the maximum entropy principle. For the profitability requirement, it may be important for online stores to improve customer lifetime value (LTV) rather than to predict future purchases accurately. We present a recommendation method for improving LTV by integrating probabilistic choice models and purchase frequency models. The proposed recommendation method finds frequent purchase patterns of high LTV users, and