LATENT PROBABILISTIC MODEL FOR CONTEXT-AWARE RECOMMENDATION

Abstract

Recommender Systems are software tools that provide personalized recommendations of relevant items to individual users. However, most of them do not take into account additional contextual information that may affect user preferences, such as place, time, or weather. Context-aware recommender systems have been proposed to solve this problem by providing the better recommendations for users based on their rating history in different situations. Since incorporating all contextual information makes the data become sparser and degrades the prediction accuracy, most context-aware methods focus on identifying and applying the relevant contextual variables into the models. However, besides the accuracy, the diversity of the recommendation is also the key to improve the users’ satisfaction on the recommended results. Moreover, most context-aware techniques have not directly considered the relationships among context, users, and items before predicting the ratings. In the real world, different contextual factors tend to affect users and items differently. This work proposes a latent probabilistic model for contextual recommendation by extending the flexible mixture model to incorporate the contextual information. Combining with the binary particle swarm optimization techniques, the relevant contextual factors to the user classes and item classes are identified and incorporated into the model. The proposed model is optimized with two cases: considering only the accuracy, and considering the trade-off between accuracy and diversity. The evaluation shows that the proposed model performs better than 1) the traditional model-based techniques that do not consider contextual information. 2) the model that considers only the relations of context to users alone or items alone, 3) the model that exploits all contextual factors, and 4) the traditional context-aware recommendation method.