Performance Comparison and Financial Distress Prediction using Machine Learning Algorithm

Predicting financial distress (FD) is essential to financial preparation, especially in the face of growing uncertainty. Small and medium-sized businesses’ intrinsic transparency makes it more difficult to make credit decisions, which raises financing costs and reduces the likelihood of acquiring funding. The use of machine learning (ML) in banking and business has advanced rapidly during the last ten years. Recent data indicate that machine learning (ML) algorithms have outperformed conventional quantitative algorithms and are now responsible for the financial sector’s remarkable transformation. The interpretability of black box ML algorithms’ estimation outcomes is a hotly contested issue. In order to comprehend the estimation findings on the database for FD prediction, this research employed SHAP scores and evaluated the forecast capability of ML methods. The outcomes suggested that the random forest and extreme gradient boosting algorithms functioned better than the others. Furthermore, according to Shapley scores, this study discovered that outcomes were significantly impacted by overall loan to equity, company price to sales, and accounts receivable due to capital. Preliminary single-period indicators were estimated using conventional financial proportions, and multi-period systems were then retrieved by augmenting them with duration, credit record, and age-related variables. Unlike previous research, FD is viewed as a major obstacle to a business’s capacity to pay its debts instead of a warning that it would fail.

Перевод пока недоступен