A New Way to Evaluate Products with Reviews: A Case Study of A Hair Dryer on Amazon Platform

Xiting Jin,Xiangdong Liu,Zhan Xiangsheng

DOI: https://doi.org/10.1088/1742-6596/1735/1/012012

2021-01-01

Journal of Physics: Conference Series

Abstract:Abstract The era of big data is coming. How to make effective use of users’ feedback information data is the key to success in various online industries. In order to help target enterprise enter the market, we build a comprehensive rating evaluation model that can evaluate the sales status, future development and contain comprehensive indicators. We firstly study the one-dimensional index. We adopt the method of correlation analysis to find the ripple effect of star rating. Then we use SPSS to draw scatter plots to analysis correlation between specific quality description and star rating level. We found through the research that “star rating”, “comment validity”, “positive and negative indicator in comments” and “member effect” are all informative for target enterprise to track, and will determine the market development of the product to a certain extent. We take advantage of the machine learning to turn the information in the text into quantifiable metrics and get polarity and subjectivity. Use NLTK for sentiment analysis of each comment to get negative, positive, neutral and compound value. Then use AHP to assign weight to the indicators wisely. With the comprehensive rating evaluation model that we build, we calculate the comprehensive rating value of all products in this industry and take the median as the benchmark value, then the products with the comprehensive rating value higher than the benchmark value can be regarded as successful products. We provide an accurate assessment. In order to effectively evaluate and predict the future sales trend of new products after their launch and effectively measure the increase and decrease of product market reputation, we establish a time-based measurement and prediction model based on Grey model theory. We can use the model to make comprehensive evaluation and effective prediction of the target company’s products. After inspection, our model can be effectively applied to different sales fields, which has great practical significance.