Abstract:

This study collected comment data from hot-selling laptops on JD.com， and used Python to perform data cleaning， text segmentation， and stop-words filtering. Afterwards， Naive Bayes， support vector machine（SVM） method， and SnowNLP method in machine learning methods are used to classify data for sentiment. By comparing the classification results of the three methods， the optimal classification method for online comment sentiment classification was selected. The LDA topic model is used to divide different topics， extract the positive sentiment word set and the negative sentiment word set in the review set， and visualize the conclusion data to further explore the highlights and drawbacks of the product. Finally， a "SnowNLP+ LDA" online sentiment analysis program based on machine learning was constructed.

Key words:

"> sentiment analysis, machine learning, SnowNLP, LDA model, SVM, Naive Bayes