The Research And Application Of Data Mining Method Based On Neural Network

With the development of Information technology, capacity of collecting andproducing data is greatly improved, corresponding to the relative lag in ability ofaccess to knowledge from Huge amounts of data. Data Mining (DM) is used toidentify effective, innovative, useful and intelligible patterns. The classification iswidely used in DM field. Construct an appropriate classifier by various methods, eg.Statistical Analysis, Machine Learning, Artificial Neural Network(ANN).ANN simulates the human brain biological neural network, making a group ofneuron nodes link together according to a certain structure, So it can deal with thevague data effectively, complex and non-linear problem in addition. Recognizablepatterns are determined by network topology structure, connection weights and nodethresholds. Objects can be optimized are network topology, weight and threshold.This paper focuses on the classifier based on BP neural network model, andoptimization algorithms, including GA, PSO and Adaboost. Compared with standardBP model, the optimized classifiers can greatly improve accuracy of classification.Further study of the Adaboost algorithm and the two types of error—FNR andFPR in classification, the paper proposed an adaptive weight algorithm based on theimportance of category, in order to improve the classification ability of the categorywith a high degree of attention.In this paper, BP neural network classifier is applied in Corporate Finance,establishing a financial crisis prewarning model of listed companies.Throughtraining sample dataset and predicting test dataset, the result verified the effectivenessand practicality of the BP classifier. By comparing the sort results ofoptimizatical BPclassifiers—Adaboost_BP classifier and improved Adboost_BP classifier, draw aconclusion that improved algorithm can make accuracy higher to some extent.