Bayesian Classification Model Based ON ISOMAP Algorithm And ITS Application

Bayes has strong mathematical theory to support and has the advantage of integrating prior information,and the classification model on the basis of this method has the advantages of interpretability,high accuracy,so they have been widely studied and applied.Under the background of China's economic transforming,Credit business in the financial industry have mushroomed,almost all financial transactions involve credit risk.When financial institutions analyze credit risk,the probability of loan default is often the focus of attention.In view of the characteristics of high-dimensional and non-linear financial data,this paper proposes Bayesian classification model based on the ISOMAP algorithm,and selects financial index data sets of 1070 listed companies involving different industries for empirical analysis,the results show that Bayesian classification model based on the ISOMAP algorithm not only optimize the model structure,but also improve the prediction accuracy in corporate credit assessment.The main work and innovation of this article are as follows:(1)In recent years,ISOMAP has greatly influenced on the study of machine learning and cognitive science,ISOMAP is gradually becoming a new research hotspot.Its essence is to find low-dimensional smooth manifolds in high-dimensional space by learning finite discrete samples,and then to find the inherent low-dimensional structure hidden in high-dimensional data,so as to achieve nonlinear dimensionality reduction or visualization of high-dimensional data.This paper constructs naive Bayesian classification model based on the ISOMAP algorithm(ISOMAP-NB)with combining the good classification effect of the Naive Bayesian classification model.(2)In practical applications,Naive Bayes assumes each attribute is independent.But In some specific circumstances,if the degree of relevance of data attributes is large,the accuracy of the classification results will be affected.considering that the attributes are not completely independent,this paper proposes Tree Augement Naive Bayesian classification model based on the ISOMAP algorithm(ISOMAP-TAN).(3)In the empirical analysis section,This paper selects the financial data of 1070 companies to test the ISOMAP-NB and ISOMAP-TAN respectively,and compares them with the Naive Bayesian and Tree Augement Naive Bayesian without data dimension reduction.Experimental result show that the two models established in this paper have good classification accuracy to some extent.