Improved Naive Bayesian Classifier for Financial Risks of Listed Companies

Abstract

Abstract In view of the deficiency of naive Bayesian classifier in the assumption of attribute independence, this paper constructs AdaBoost-naive Bayesian classification model to improve the accuracy of the classifier through continuous machine learning. Through data simulation, it is found that with the increase of sample size, the fluctuation gradually decreases, the accuracy reaches more than 99%, and the trend is stable. When the sample attribute is less than 400, the correct rate of the model reaches more than 95%, and the trend is stable. When the sample attribute is more than 600, the correct rate decreases to about 50%. The fewer classification categories, the higher the correct rate of the model. When the number of classification categories is more than 50, the correct rate is zero. In the empirical analysis on the financial risk rating of listed companies in the cultural industry, the improved naive Bayesian classification algorithm has significantly higher efficiency than naive Bayesian classification algorithm, and the model is more sensitive to samples with higher financial risk. The empirical analysis shows that the improved naive Bayesian classifier has higher accuracy and reliability. Through robustness analysis, it is also found that the improved naive Bayesian model has strong robustness.