Research On Dynamic Modeling Of Financial Crisis Warning Based On Resampled CBR

The rapid development of economy makes the current market environment more complex,the financial risks faced by enterprises more diversified,and the uncertainties faced by investors,creditors and other stakeholders are higher.As manufacturing industry accounts for the highest proportion of quantity and amount in all industries,the stability and health of its development has a vital impact on the whole market economy.Therefore,we should Emphasis should be placed on the financial risk early warning of manufacturing companies.That is,how to construct an efficient early warning model of financial crisis through the internal and external indicators disclosed by enterprises,predict the current and future financial situation of enterprises,warn the enterprises with financial crisis,and realize the good operation of enterprises is an urgent problem to be solved at present.Combining with the research of experts and scholars at home and abroad,we find that:(1)in the previous research on financial crisis early warning,the number of ST companies and non-ST companies in the data set is mostly in a balanced state or a specific proportion,without considering the performance of the early warning model under other proportions;(2)In the current research on enterprise financial crisis early warning,the construction of early warning index system is usually only considered.Considering the financial indicators and ignoring the non-financial indicators;(3)Based on the concept drift(CBR),the model constructed with CBR as the basic classifier ignores the natural characteristic that the new samples have more reference value than the old samples,tliat is,the influence of sample time weight on the performance of the model is not considered.Based on the financial data and non-financial data of 11146 A-share manufacturing listed companies from 2007 to 2017,this paper studies the dynamic modeling of enterprise financial crisis early warning.Firstlyr two types of samples in sample companies are resampled to compare the effect of resampling on the performance of the model.Secondly,considering the different importance of the new and old samples,this paper constructs a CBR model considering the time weight of samples,and compares the results with those withut considering the time weight.Finally,considering the dynamic model versus the static model In this paper,a dynamic CBR model considering concept drift and a dynamic CBR model considering both concept drift and sample time weights are constructed,and its performance is compared with that of the static CBR model.Empirical findings are as follows:(1)Improving the proportion of minority samples in total samples by resampling can significantly improve the prediction accuracy of the model for minority samples,especially when the number of two types of samples reaches equilibrium,the prediction accuracy of each model for minority samples reaches the highest level;(2)The prediction accuracy of the model considering the time weight of samples is guaranteed while the overall prediction accuracy is guaranteed;(3)The prediction accuracy of dynamic early warning model is better than that of static early warning model.In conclusion,this paper studies the influence of sample time weights on the performance of static model and dynamic model,provides a new idea for the research of dynamic model of financial crisis early warning,and provides a new feasible method for the financial situation prediction of manufacturing companies.