Study On Text Classification Method Based On Adaptive Genetic BP Neural Network

In this paper, the text classification of the status quo and problems of development were systematically described by the general process of text categorization text classification-related technologies are introduced and discussed. Focus on analysis and study of the text that the feature selection algorithm for text classification technology, and key technologies such as text classification.This is more systematically reviewed and studied the theory of genetic algorithms, BP neural network algorithm the basic principles of the adaptive genetic algorithm, respectively, BP neural network learning and training and classification algorithms are briefly discussed.Combination of genetic algorithm and the respective advantages of BP neural network is proposed based on adaptive genetic algorithm and BP neural networks combination of text classification methods, although the BP neural network compared to other algorithms have stability, strong anti-interference, etc., the more suitable for text classification, but there are learning efficiency is low, slow convergence and easy to fall into local minimum points of disadvantages. Therefore, this article will introduce the genetic algorithm BP neural network, the establishment of GA_BPNN neural network model.But the genetic algorithm in the evolutionary process, because random variation and cross-cutting, no purpose, leading to low efficiency of the algorithm, so this paper introduces the self-adaptive genetic operators to control the direction of the evolution of populations, while for the repeated evolution of genetic algorithm, the easy to tend to a fixed direction of evolution, the paper simultaneously with multiple groups of genetic manipulation, to maintain population diversity is conducive to maintaining an excellent group of individuals. The improved genetic algorithm and BP neural network algorithm has been combined with related algorithms are compared and analyzed, theoretically proved that the improved algorithm, the algorithm's time complexity is superior to similar algorithms, to a certain extent to improve the text classification accuracy, while the algorithm model can be a very good handle non-normal text belong to many types of problems.And accordingly design and implement a combination of adaptive genetic algorithm and BP neural network algorithm for text classification experimental system, compared to the Naive Bayes algorithm and the traditional BP neural network algorithm. Experimental results show that using this method of text classification algorithm achieved good classification results.Finally, the achievements and insufficient points of the article were concluded, the next research was looked ahead.