Mobile devices, such as cellular phones and car navigation systems, are essential to daily life. People acquire necessary information and preferred content over communication networks anywhere, anytime. However, usability issues arise from the simplicity of user interfaces themselves. Thus, a recommendation of content that is adapted to a user's preference and situation will help the user select content. In this paper, we describe a method to realize such a system using Bayesian networks. This user-adapted mobile system is based on a user model that provides recommendation of content (i.e., restaurants, shops, and music that are suitable to the user and situation) and that learns incrementally based on accumulated usage history data. However, sufficient samples are not always guaranteed, since a user model would require combined dependency among users, situations, and contents. Therefore, we propose the LK method for modeling, which complements incomplete and insufficient samples using knowledge data, and CPT incremental learning for adaptation based on a small number of samples. In order to evaluate the methods proposed, we applied them to restaurant recommendations made on car navigation systems. The evaluation results confirmed that our model based on the LK method can be expected to provide better generalization performance than that of the conventional method. Furthermore, our system would require much less operation than current car navigation systems from the beginning of use. Our evaluation results also indicate that learning a user's individual preference through CPT incremental learning would be beneficial to many users, even with only a few samples. As a result, we have developed the technology of a system that becomes more adapted to a user the more it is used.

This paper proposes a novel approach of constructing statistical preference models for context-aware personalized applications such as recommender systems. In constructing context-aware statistical preference models, one of the most important but difficult problems is acquiring a large amount of training data in various contexts/situations. In particular, some situations require a heavy workload to set them up or to collect subjects capable of answering the inquiries under those situations. Because of this difficulty, it is usually done to simply collect a small amount of data in a real situation, or to collect a large amount of data in a supposed situation, i.e., a situation that the subject pretends that he is in the specific situation to answer inquiries. However, both approaches have problems. As for the former approach, the performance of the constructed preference model is likely to be poor because the amount of data is small. For the latter approach, the data acquired in the supposed situation may differ from that acquired in the real situation. Nevertheless, the difference has not been taken seriously in existing researches. In this paper we propose methods of obtaining a better preference model by integrating a small amount of real situation data with a large amount of supposed situation data. The methods are evaluated using data regarding food preferences. The experimental results show that the precision of the preference model can be improved significantly.