A Movie Genre Prediction Based on Multi-Variate Bernoulli Model and Genre Correlations

Abstract

In this paper, a movie category based on Bayesian model and categories correlations is proposed. Although several methods have been reported on improving the user satisfaction based on unexpectedness metric, to the best of our knowledge, predicting items’ categories rather than predicting items’ rating is a new attempt. This in turn completes the items’ categories given by experts and improves user satisfaction by providing a surprise effect in the recommendations given to users. We employ Bernoulli multivariate model to estimate a likelihood of a movie given category and the Bayes rule to evaluate the posterior probability of a genre given a movie. Experiments with the MovieLens dataset validate our approach.