A Logistic Regression Approach for Generating Movies Reputation Based on Mining User Reviews

Abstract

The paper aims to present an approach for generating a single reputation value towards a target movie based on mining movie reviews and their attached ratings with the use of Logistic Regression classifier and Latent Semantic Indexing (LSI) method. The contribution of the paper is fourfold. First, we apply Logistic Regression classifier to determine the sentiment orientation of movie reviews (positive or negative). Second, we use LSI method and cosine similarity to compute the semantic similarity between reviews. Third, we compute a custom reputation value separately for positive opinions group and negative opinions group. Finally, we use the weighted arithmetic mean to generate a single reputation value towards the target movie.