Predicting Online Review Scores Across Reviewer Categories

Abstract

In this paper, we propose and test an approach based on regression models, to predict the review score of an item, across different reviewer categories. The analysis is based on a public dataset with more than 2.5 million hotel reviews, belonging to five specific reviewers’ categories. We first compute the relation between the average scores associated with the different categories and generate the corresponding regression model. Then, the extracted model is used for prediction: given the average score of a hotel according to a reviewer category, it predicts the average score associated with another category.