Abstract—Consumer reviews and ratings of a product or service on websites is an important factor in a potential customer’s decision-making process. The ratings given by the consumers or reviewers are also valuable to the product vendor or service provider for market segmentation or product and service evaluation and improvement. Since the ratings can be treated as discrete ordinal data, we apply the Combination of Uniform and shifted Binomial mixture (CUB) model to analyze review ratings. The CUB model gives an interesting perspective on review ratings modeling and analysis since the model provides a tool to interpret the ratings in terms of the reviewer’s level of feelings and uncertainty towards the product or service. The CUB model is also able to incorporate covariates, such as the reviewer’s gender, to explain the level of feelings and uncertainty. We illustrate with examples from two well-known datasets in the field of data mining.
Index Terms—Mixture models, ordinal data, review ratings, statistical models.
Y. C. Low is with the Department of Computing and Information Systems, Sunway University, Bandar Sunway, 47500 Subang Jaya, Selangor, Malaysia (e-mail: yehchingl@sunway.edu.my).
[PDF]
Cite: Yeh Ching Low, "Statistical Modeling for Review Ratings Data," International Journal of Knowledge Engineering vol. 3, no. 2, pp. 48-51, 2017.
