Sentiment classification of micro-blog comments based on Randomforest algorithm

摘要

Sentiment classificationofMicro-blogcommentsaims to distinguish consumers' attitudes towarda certain brand or an event. A great number of research studies related to sentiment analysishave been conducted based on machine learning methods, such as Support Vector Machine andNaive Bayes, to build the classifier, and some other research studies tried to build one based onemotional thesaurus. However, the rate of accuracy is yet to improve. In order to improve theaccuracy rate further, this thesis bases on the Randomforest algorithm to construct an emotionaltendency classifier to help the mobile phone service providers to further understand consumers'attitudes toward their brand. TheRandomforest algorithm is a bagging algorithm of the ensemblelearning with some weak classifiers,which works verywell in anti-noise and reducing the overfittingproblem. Finally, a classifier with an 83% accuracy rate for sentiment classification ofmobilephone brands on Micro-blog comments was built based on Randomforest by using R languagesoftware.