Topic Model based Ensemble Learning for Rating Prediction and Its Application
Liwei Shao, Lei La, En Zhang, Honglei Hua · 2021
User rating prediction is an important problem in recommendation system, this task has been plagued by data sparsity for a long time. Especially, the rating prediction based on product features, user behavior and other background data is not only troubled by cold start, but also difficult to find the migration of user interest. In order to solve above problems, this paper proposes an integrated learning method based on topic model. In this method, the interest points of user comments are obtained by topic model questions, and then the score is predicted based on the improved CatBoost. Through this strategy, we can use the semantic information of the comment text more effectively for rating prediction. The proposed method is implemented in an online travel data analysis platform, experiment result shows it has ideal performance in rating prediction of online travel community.