Nonstationary and Imbalanced Dataset for Financial Distress Prediction

Financial distress identification is crucial for companies, investors, and authorities. This dataset provides real-world indicators of Brazilian enterprises based on data from the Brazilian Securities and Exchange Commission (CVM), organized quarterly in a non-stationary manner. It covers 13 years (2011 to 2023) and includes 1,004 different corporations and 32,005 records, each described with 84 indicators. The data are strongly unbalanced: 2.76% are data of companies in financial distress situation, while 97.24% are not.