Abstract: 【Purpose/significance】Sentiment analysis based on massive internet comment data, mining the key factors that affect cus⁃ tomer service evaluation and satisfaction, and understanding the reasons behind negative reviews are of great significance to improving the level of enterprise customer relationship management.【Method/process】Through word cloud, feature association analysis of the se⁃ mantic network, the feature analysis of the LDA topic model, and the semantic-based sentiment dictionary method, we performed senti⁃ ment calculation and analysis based on the customer reviews data of ten express companies in Baidu reputation.【Result/conclusion】 The main factors affecting the emotional tendency of customers include: logistics speed, service attitude, telephone service, complaint handling, logistics information update, timeliness and charging price. Based on the above results, countermeasures and suggestions are proposed.【Innovation/limitation】Based on the real world data, the method of data mining is used to analyze customer emotion tenden⁃ cy, which provides a research paradigm of data science for the identification of key influencing factors of customer emotion.