MVLSC Home · Issue Contents · Forthcoming Papers
Beyond Traditional Ratios: Predicting Financial Distress through Compositional Data and Fuzzy Clustering in Agricultural Retail
Xavier Molas-Colomer, Joan Carles Ferrer-Comalat, Marc Carreras-Pijuan and Salvador Linares-Mustarós
Bankruptcy prediction plays a crucial role in financial risk assessment, but traditional methods based on classic financial ratios face important limitations due to asymmetry, outliers, and redundancy. This study explores whether combining Compositional Data Analysis (CoDa) with fuzzy cluster analysis provides a more reliable approach to bankruptcy prediction. Using financial data from agricultural retail firms obtained from the SABI data set with last available financial year 2024, or at least 2023, financial statements are transformed into centered log-ratios and analyzed through fuzzy clustering to identify distinct financial profiles. The resulting cluster membership values, treated as compositional data, are transformed into additive log-ratios and used as predictors in a logit regression model to estimate bankruptcy probabilities. The results indicate that the compositional approach can improve prediction accuracy by adequately addressing the relative nature of financial ratios while reducing the distorting effects of asymmetry and outliers, improving conventional methods.
Keywords: Compositional data analysis (CoDa), financial ratio, cluster analysis,fuzzy cluster, bankruptcy prediction