Abstract: In order to explore the important influence of online comment data of agricultural products on consumer satisfaction decision-making, the selenium-enriched tea with experience was selected as the research object, machine learning classification algorithm was used to classify the sentiment of the selenium-enriched tea review text, and TF-IDF and LDA models perform visual mining of text feature words and topics, and identify factors affecting consumers’ satisfaction with selenium-enriched tea. The study found that consumer satisfaction was mainly embodied in the five factors of product trust perception, marketing perception, quality perception, logistics service and price perception. Quality perception and trust perception were the main factors that affect negative consumer satisfaction ratings. Through machine learning and data mining, different dimensions of consumer satisfaction indicators for selenium-enriched tea are identified, which can provide marketing decision-making reference for enterprises and governments, and provide reference value for promoting the brand building and industrial development of selenium-enriched tea of regional characteristic agricultural products.

Key words: machine learning, sentiment classification, LDA, TF-IDF, satisfaction perception