Predictive Modeling in Finance: Harnessing Machine Learning Algorithms for Enhanced Decision Making

Abstract

Predicting financial stability is crucial for assessing risk and making informed decisions in the financial sector. Accurate predictions can help prevent financial crises and guide strategic planning for companies and investors. Various machine learning algorithms have been employed to enhance prediction accuracy for economic distress, including XGB, LGBM, Linear Discriminant Analysis, and Logistic Regression. These models were assessed based on key performance metrics: Accuracy, ROC AUC, and F1 Score. The result revealed that LDA excels with an ROC AUC of 0.90 and an F1 Score of 0.98, demonstrating its superior ability to balance precision and recall and effectively differentiate between distressed and non-distressed entities. While the XGB Classifier and LGBM Classifier also show strong performance, they do not exceed LDA in overall effectiveness. These results highlight the importance of leveraging multiple evaluation metrics to select the most suitable model, with LDA emerging as the most reliable choice for accurate financial distress predictions.