A new recommendation framework for accommodations sharing based on user preference

Abstract

The population of accommodations sharing makes it necessary to make personalized recommendations for users. But general methods perform poorly when applied to accommodations sharing due to severer sparse data and user stickiness. In order to conquer these gaps, this study proposes a specific framework with considering the information of user preference through LDA and Naive Bayes Method. These two methods enable to transfer the user preference into quantitative analysis. Furthermore, LFM possesses the function that can forecast the missing parts through learning user-item rating matrices. It generates the first recommendation list based on the quantized characterization of user preference. The second recommendation list can produce by the Naive Bayes Method. Consequently, the final optimal recommendation result of accommodation sharing is gathered from combining two recommendation lists. Finally, experiments based on the Airbnb real dataset demonstrate the promising potential of this study.