Abstract:
"This study investigates the application of Machine Learning techniques in predicting the financial distress of listed Licensed Finance Companies (LFCs) operating in Sri Lanka. A variety of Machine Learning models including Logistic Regression, Decision Tree, Support Vector Machine, Naïve Bayes, XGBoost, Neural Networks, and Random Forest were evaluated for their predictive power, on data pertaining to key financial soundness indicators of LFCs categorized under the CAMEL Framework spanning over the past 13 years from 2010-2022. The study demonstrated that advanced models such as XGBoost and Random Forests outperformed the simple and traditional techniques in terms of accuracy and predictive power.
The analysis included a feature importance assessment to provide insight into the key financial ratios contributing to financial distress prediction and aiding in providing actionable recommendations for the regulatory entities based on the prediction outcomes of the best-performing models.
Furthermore, the study underscores the importance of timely and accurate predictions in the financial sector, as early detection of distress can help mitigate systemic risks. The findings suggest that incorporating ML tools can enhance regulatory oversight by offering precise, data-driven recommendations for policymakers.
Despite the promising results, the study acknowledged several limitations regarding data availability and model interpretability. This research contributes to the literature on financial risk management by highlighting the potential of integrating Machine Learning techniques in the financial domain to promptly detect financial distress among the LFC sector paving the way to avoid potential damage to the financial system and ultimately to the public.
"
