The application of PROMETHEE multi-criteria decision aid in financial decision making: case of distress prediction models evaluation

Abstract

Conflicting rankings corresponding to alternative performance criteria and measures are mostly reported in the mono-criterion evaluation of competing distress prediction models (DPMs). To overcome this issue, this study extends the application of the expert system to corporate credit risk and distress prediction through proposing a Multi-criteria Decision Aid (MCDA), namely PROMETHEE II, which provides a multi-criteria evaluation of competing DPMs. In addition, using data on Chinese firms listed on Shanghai and Shenzhen stock exchanges, we perform an exhaustive comparative analysis of the most popular DPMs; namely, statistical, artificial intelligence and machine learning models under both mono-criterion and multi-criteria frameworks. Further, we address two prevailing research questions; namely, "which DPM performs better in predicting distress?" and "will training models with corporate governance indicators (CGIs) enhance the performance of models?”; and discuss our findings. Our multi-criteria ranking suggests that non-parametric DPMs outperform parametric ones, where random forest and bagging CART are among the best machine learning DPMs. Further, models fed with CGIs as features outperform those fed without CGIs.