Discovering the Trends in the Density of Reviews for Different Products an Intelligent Data-Driven Method

Abstract

The density of reviews is a metric that measures the population’s actual propensity to write a review. Knowing the density of reviews could assist retailers in strategically adjusting their marketing efforts with regard to writing reviews. This paper proposed a novel data-driven framework using genetic programming (GP) to intelligently model the relationship between the number of reviews and the number of products purchased for different product categories. Without any prior assumptions, we automatically learned the structure and parameters of the most feasible relationship models for 14,956 product items in 83 product categories. We find that review density is a function of purchase quantity in the majority of project categories. In particular, the review density of search products will be likely to decrease, while the review density of experience products will be likely to increase. Our findings provide important implications for methods of relationship model building and electronic commerce (eCommerce) marketing. Evaluations show that the proposed approach can intelligently learn the generalized knowledge beneath a large amount of data for marketing decision, and the model written in the form of mathematical functions with high accuracy and interpretability.

Keywords: Electronic commerce, Online reviews, Genetic programming, Artificial Intelligence