The Influence of Data Imputation Methods on the Classification Efficiency of the Logit Model Used for Forecasting the Bankruptcy of Companies

Forecasting the bankruptcy of companies exposes the missing data problem, which applies chiefly to entities having financial problems, who wish to conceal thereby their bad situation. One of the methods of making up incomplete data is imputation. The aim of the paper is to present different data imputation variants and to compare their influence on the classification efficiency of one of the statistical bankruptcy forecasting methods – the logit model. The results have shown that the best approach is to use the median as determined separately for healthy and bankrupt companies.