The invention discloses a merchant comment analysis system based on machine learning. The merchant comment analysis system comprises a data processing layer, a machine learning algorithm layer and a user service layer, wherein the data processing layer is used for acquiring platform data, performing preprocessing, text word segmentation and word cloud generation on the platform data, and outputting an evaluation data set; the machine learning algorithm layer is used for performing model training, sentiment analysis and clustering analysis by using a comment data set, and outputting the positive sentiment degree and merchant classification labels of the comment data; and the user service layer provides merchant recommendation, search, classification, merchant data analysis and comment data analysis information for users. According to the invention, a machine learning algorithm is adopted to analyze the comment data of users, quantify the sentiment data of the users and mine the emotion and attitude of the users in the comments, so merchant evaluation can be refined, the shortages of platform scoring are made up, a multi-dimensional scoring model is constructed, and a more accurate merchant evaluation system is realized.