Document-Level Sentiment Classification Based on Behavior-Knowledge Space Method

There are mainly two kinds of methods for document-level sentiment classification, unsupervised learning and supervised learning. When ensemble learning is introduced, existing methods only combine unsupervised learning algorithms or supervised learning algorithms. To overcome each other’s flaws, a novel sentiment classification method based on behavior-knowledge space is proposed, in which two unsupervised and two supervised learning algorithms are utilized. The experiment results not only explain the effectiveness by diversity measure, but also show that the proposed method is significantly superior to each individual classifier. In addition, our method is better than the other two common ensemble methods.