Applying machine learning methods to analyze customer comments about fresh food on e-commerce platforms in Vietnam

Abstract

In recent years, farmers have developed the sale of agricultural products toward consumers via e-commerce platforms. E-commerce has become a new and effective way to help farmers access the market. Thus, in comparison to other commodities, agricultural products are heavily affected by seasonality, with complex factors such as short shelf life, vulnerability to damage, and high transportation costs. Consumers set high standards for the quality, speed of delivery, frequency of consumption, and unit price of these products. Analyzing customer reviews helps businesses discover consumer decision-making mechanisms, thereby forming an appropriate marketing strategy for their agricultural products. Besides, they will see what customers are unsatisfied with to solve and improve the quality of products and services. In this study, the authors research and propose machine research methods to classify and screen customers' comments about agricultural products on three e-commerce platforms: Tiki, Sendo and Voso. Experimenting with the model on the collected data set with the results of the sgdclassifier algorithm combined with the One-vs-Rest method gave the best prediction results with 87%. The study also builds charts and directly shows the amount of data analyzing the factors affecting customer satisfaction with quality products as well as seller's services and e-commerce platforms. In addition, the study proposes recommendations to help businesses improve the quality of products and services, thereby providing better strategies to attract and retain customers.