Movie Prediction Based on Feature Generation

Social media provide people with amazing platforms to improve their experiments and offer interesting items. Predicting which movies users like or not is solved by several techniques; however, it is still a challenging problem. To address the problem of movie prediction, we propose a method to predict whether movies viewers like or not based on an analysis of movie ratings and user profiles. First, we consider each user as one data table, including features and a class label (Like or Dislike). Then, we select four machine learning techniques to classify the data table of the user. Experiments were conducted on the real movielens data. We selected 3000 users (data tables) from the movie data. Each data table had 26 features and at least 80 samples (movies). The empirical results show that bagging model predicts more accuracy than other algorithms. Moreover, this method can be implemented as an online tool to recommend appropriate films to viewers.