The construction method of the financing constraint index of enterprises generally adopts a two-step method.First,it is judged whether enterprises has financing constraints,and then the financing con-straint index is constructed through financial indicators.This paper uses a machine learning approach to ex-plore the construction of an enterprise's financing constraint index.Firstly,in determining whether an en-terprise has financing constraints,a clustering algorithm(K-mean clustering and systematic clustering)is used to classify these enterprises,and then on the basis of clustering,classification algorithms(decision tree,logistic regression and neural network)are used to explore the impact characteristics of the enter-prise's financing constraints.The results predicted by the classification algorithms yielded the best accu-racy of the decision tree algorithm,the best stability of the logistic regression model predictions and the poorer neural network prediction model.The probability of financing constraints predicted using the logis-tic regression model is more closely related to the SA index and Zhu Xuewen's LFC financing constraints in-dex.In terms of the important characteristics of a firm's financing constraints,the important characteris-tic variables in the pre-classification of clustering are interest cover multiple and dividend payout ratio,and in the algorithm of classification,the important characteristic variables are financial expense ratio and net asset payout ratio.
万方数据