Study On Text Classification Based On Multi-classifier Fusion

The rapid development of socio-economic and information technology leads to the explosive growth of various information, but the text data is still the most important and the most direct information carrier. In order to extract valuable information from the mass text information quickly and efficiently, analysis method and utilization method of data has shown their necessity, which has maken information retrieval, information filtering technology being further studied and widely used, from which the step of text classification is indispensable. As the basis for the above application technology, text classification especially Chinese text classification technology, one key branch of text classification, possesses important significance.It is not an easy task for computer to determine the type of the text automatically. Category standard needs to be set up in advance. Text needs to be described in symbolic and be converted into the mode that computer can recognize and understand. Then, corresponding text classification algorithms should be designed. Thus, take advantage of computer powerful computing capability instead of artificial classification.This topic studies on Chinese text classification based on multi-classifier fusion using fuzzy integral. Firstly, explains the relevant background and significance, analyses the research status of domestic and international text classification technology and narrates the research methods and content used in this paper. Secondly, introduces technology of neural network classifier and k NN classifier, and considering that different Chinese text classification methods have advantages and disadvantages respectively, so put forward a new opinion that it can further improve the accuracy of classification by multi-classifier combination, and the fuzzy integration is one of the effective combination methods. Thirdly, around Chinese text classification based on multi-neural network classifier fusion and Chinese text classification based on multi-k NN classifier fusion, which use fuzzy integral as fusion tool, carries out a detailed exposition from model to the application in text classification, and discusses the applicable situations of these two Chinese text classification methods based on multi-classifier fusion.As the main research work, this paper combines BP neural network classifier, RBF neural network classifier and RBF neural network classifier with K-means algorithm by using Sugeno fuzzy integral and Choquet fuzzy integral as fusion operator. Then similarly, combines k NN classifier, within-class average value k NN classifier and improved k NN classifier based on the central vector classification by using Sugeno fuzzy integral and Choquet fuzzy integral as fusion operator, in order to achieve more ideal text classification results.Finally, compares the calculation results of each single text classifier and text classification methods on multi-classifier fusion by using example, which show that the classification results of multi-classifier fusion text classification methods based on fuzzy integral have a certain improvement over the traditional single text classifier in accuracy. Explore a new idea for Chinese text classification algorithm optimization, through the study of this subject.