How to extract effective information that affects consumer satisfaction from online comments has become a hot issue for customer behavior. This article is based on the data mining of online comments and the research object are the top-selling tablets on the JD platform from October to December 2018. We started by analyzing influential factors such as goods, after-sales service, and logistics, and crawled online review information of nearly 3,000 tablet computers from five major brands. We first use the jieba word segmentation tool to process the user comments, and use TF-IDF to calculate the frequency of different words in the comments to determine the main keywords of the comments. Secondly, we set up a user's sentiment dictionary to determine the sentiment index of the review, and combined the keywords and sentiment index to get the degree of consumer satisfaction on different influencing factors. Finally, we imported the quantified characteristic factors into Clementine 12.0, and established a Bayesian network model of customer satisfaction, thereby obtaining a ranking table of the importance of each factor to product sales. To improve the model robustness, we adopt a multivariate linear model to check the accuracy of the output results. Our research can not only formulate effective product service sales strategies for merchants, but also guarantee customers to experience better products and services.