Research On The Emotional Tendency Of Micro-blog Based On Text Classification

In recent years, more and more people rely on the use of some social software to express their views, such as micro-blog, forum, post bar and other channels to express their views, expressing emotions. And people for the hot events and topics of a variety of emotional information, through the above way intuitive reflected. Therefore, through the mining and study of the public’s emotional tendency to all kinds of events, it can achieve the perception of hot public opinion and public opinion, and provide important basis for decision-making. Previous information retrieval and data acquisition technology, most are based on keywords, it is difficult to support the sentiment orientation of mining, information extraction and text classification usually without deep semantic mining and text in the expressed emotion tendency of unable to deep mining. Therefore in the era of big data and effectively using the related knowledge of data mining and text mining, mining hot microblogging information and comment on the sentiment orientation, in intelligent commodity recommendation, government public opinion monitoring, automatic text classification has the broad prospects for development.The main research contents of this thesis are as follows:(1) This paper introduces the related concepts of sentiment analysis and text classification and analysis of micro-blog technology, short text sentiment classification and text classification research status at home and abroad, introduces several commonly used text classification algorithm, support vector machine(SVM), maximum entropy, decision tree, artificial neural network algorithm. And introduces the traditional K nearest neighbor algorithm, the basic idea of the algorithm and the application of the algorithm in the text classification.(2) The traditional k nearest neighbor algorithm based on association rules, and combined with the proposed combination k nearest neighbor algorithm based on correlation in the emotional theme, theme and emotion associated text classification, combination of characteristic value, better able to determine the value of K, improve the classification efficiency.(3) This paper will be improved after the theme of emotional K recent neighbor algorithm in hot microblogging information data concentrated application and emotional tendency of positive, negative and neutral classification experiments based on and the traditional K recent neighbor algorithm by comparative experiments. Micro-blog short text preprocessing, feature of frequent itemsets to establish other work before the experiment. Experimental results show that compared with the traditional K recently neighbor algorithm, the improved algorithm in precision, recall, and F1 value in the evaluation index are improved, reducing the time complexity of the algorithm, and improve the efficiency of sentiment classification, in favor of more accurate microblog hot information in sentiment, has certain feasibility in dealing with massive contains the theme of the popular Chinese microblogging.