Data
Sentiment classification is a hot research direction at present, but most research is based on balanced data sets. In real life, the sample is impossible to balance. For sentiment analysis of unbalanced data, we not only need to pay attention to the overall classification performance, but also need to care about the classification performance of a few classes. How to improve the recognition rate of a few types of samples while improving the overall recognition rate has become a research hotspot. Aiming at this problem, this paper proposes a model based on ensemble learning, extracts features by TF-IDF+SVD, and integrates five base classifiers by stacking to sentiment classification. The experimental results show that it can be more effective in emotional classification on unbalanced data sets than other methods.
