A Hybrid Network Analysis and Machine Learning Model for Enhanced Financial Distress Prediction

Abstract

Financial distress prediction has become increasingly critical in the current economic climate, where timely identification of at-risk companies can prevent substantial financial losses and systemic failures. This study proposes a novel hybrid model that integrates network analysis with machine learning techniques to enhance the prediction of financial distress. The model utilizes the “Financial Distress” dataset from Kaggle, which classifies companies based on their financial health using a defined threshold: companies with a distress score below or equal to -0.50 are considered financially distressed, while others are deemed healthy .To improve predictive performance and reduce overfitting, ensemble learning techniques such as Random Forest and Voting Classifier are employed. These models aggregate predictions from multiple base learners, leveraging their combined strengths to enhance accuracy and robustness. Additionally, network analysis is used to uncover hidden patterns and relationships among financial entities, providing a systemic view of financial risk. The hybrid approach offers a comprehensive solution that captures both individual company metrics and broader financial interdependencies. This system is intended to serve as a powerful decision-support tool for analysts, regulators, and stakeholders, enabling early intervention and better risk management in financial ecosystems.