A Hotel Hybrid Recommendation Method Based On Context-Driven Using Latent Dirichlet Allocation

Abstract

Recommender systems play an important role in helping users find items that they want. Normally, ratings are used in content-based filtering (CBF) and collaborative filtering (CF) for recommendation. However, only ratings are not enough for recommendation. Thus, contextual information, context driven and Latent Dirichlet Allocation (LDA) are used to improve recommendation. Also, the context of individual user has changed in the timeline (context-driven). In this work, a hotel hybrid recommendation method (CF+CBF) based on context-driven using LDA is proposed. Firstly, we find missing user ratings of user-hotel rating matrix by applying LDA on user ratings in order to get predicted score of hotels for the target user. Secondly, we find a group of users similar to the target user (neighbors). Then, we apply context-driven to recommend hotels that meet current interest of the target user. To evaluate the proposed method, we compare our proposed methods either CBF or CF integrating with LDA by measuring nDCG. The result shows that our proposed method outperforms all comparable methods in result accuracy.