Research on the relationship between e-commerce ratings and reviews based on Naive Bayes

Abstract

With the popularization of the Internet, online shopping has gradually replaced traditional shopping methods. Product ratings and comments from buyers have become important indicators for users when purchasing products. Because of this, the behavior of employing someone to rate high, known as click farming, has gradually risen. To deal with this phenomenon, users pay more attention to the content of the review itself rather than the score to make judgments. Then why are the stores still keen on employing click farm? We think that the high scores seem to play a guiding role in the user's comments. To explore whether this phenomenon exists, we crawled sales information from Amazon and built a mathematical model. A Naive Bayes Classifier is employed to screen out the high-frequency vocabulary from the comments. And we defined the important concept of Irrational Discrimination base on Multiple Linear Regression. After analyzing the data for the whole year, we found that discrimination does exist. The specific vocabulary appearing in the review has a large star rating relevance shows that not only the time guiding role exists between star ratings and reviews, but also a strong interaction.