DbRMP: Predicting Douban Rating of Movies with high-dimensional Features by Comprehensive Machine Learning Algorithms

Abstract

Nowadays, as the motion picture industry has become a multi-billion dollar business, it is not only a centre of entertainment. Movie investors normally value the rating of movies as utmost important since a bad rating can directly discourage people from watching the film and lead to a failure of investment. As such, the prediction of movie rating is essential to the film investors and companies for avoiding investment risks. In this paper, we propose a machine learning based method, called DbRMP, to find the optimal machine learning model for predicting the rating of movie in Douban (The largest online database of movies in China). Our method is based on different attributes / features of the movie obtained from Wikipedia, Douban, Baidu Baike, IMDd. We propose a data augmentation method to expand the size of the dataset. In the experimental evaluation, traditional machine learning models and deep learning models are investigated using this dataset. In the traditional ML evaluation methods, the experimental results show that GBDT outperforms other considered machine learning models by achieving an accuracy of 80.29% on the test set; in the deep learning methods, CNN offers the best accuracy, f1-score, precision and recall, which are 77.85%, 78.33%, 80.38% and 73.31%, respectively.