The Research And Implementation Of Sentiment Classification Algorithm For Micro-blog

Recently, with rapid development of micro-blog, how to use vast amounts of data mining useful information becoming a hot topic. Micro-blog emotion classification is very popular, which has a wide range of applications, such as:analysis and prediction of current events, feedback of commodities and product reviews, analysis of consumer preferences and monitoring public opinion in network. Micro-blog sentiment classification study is divided into three areas:emotion based dictionary, machine learning and feature fusion. Because micro-blog has little information, freedom of language style, a lot of noise, so micro-blog sentiment classification research does not have a satisfactory result. In response to this situation, this paper proposes three options to improve the results of micro-blog sentiment classification. The main contribution of this paper are shown as follows:First, this paper proposes a sentiment classification algorithm based on the emotional affective computing. The targets of the algorithm is to determine the polarity of micro-blog, which contains significant positive and negative emotion words. This algorithm combines four common emotional dictionaries, which describes the micro-blog from multiple dimensions. After calculating micro-blog emotional tendencies based on the calculating of emotional word, the category of the micro-blog is determined.Second, this paper proposes a sentiment classification algorithm based on classifier fusion. The algorithm is suitable for large amounts of data. The algorithm analyses the emotion classification method with machine learning. It extracts three different kinds of classifier prediction labels and prediction scores; Then, it puts the labels and scores as a new feature to train; Finally, it predicts the category of the micro-blog.Third, this paper proposes a sentiment classification algorithm based on emotion feature fusion. The algorithm combines the human emotion analysis and machine learning method. Depending on the analysis of the characteristics of micro-blog and the expression of human emotions, it selects four characteristics; Then, it uses an objective statistical features, such as the commonly used chi-square characteristics; Finally, it generates a new feature by fusing features, SVM classifier is used to training model and predict the category of the micro-blog. Experimental results demonstrate that the proposed methods can effectively improve the results of micro-blog sentiment classification.