Machine learning models for optimal DEBT capital structuring in high-growth renewable energy firms

This paper investigates the use of machine learning (ML) techniques to determine optimal debt capital structures for high-growth renewable energy firms operating in complex and uncertain financing environments. Drawing on a simulated panel dataset that mirrors firm scale, growth dynamics, project risk profiles, credit quality, and macroeconomic conditions, the study applies and compares three modeling approaches linear regression, random forest, and gradient boosting to predict both optimal debt ratios and long-term debt maturity compositions. The analysis integrates permutation-based interpretability methods to ensure transparency and explain the contribution of individual financial and risk variables to model outcomes. Empirical results demonstrate that ensemble learning models substantially outperform the linear benchmark in both predictive accuracy and explanatory power, highlighting their ability to capture nonlinear and interaction effects among firm-level and market variables. Profitability, volatility, project risk, and credit rating emerge as the most influential determinants of capital structure, confirming theoretical expectations from trade-off and pecking order perspectives while revealing more complex patterns than traditional models can detect.