Scholarlink

<p>This study proposes an interpretable machine learning model to predict a firm’s profitability in the context of post-COVID India, based on the previous year’s financial ratios. The dependent variables include Return on Equity (ROE) and Return on Assets (ROA), which are modelled one at a time to investigate whether pre-target financial fundamentals are predictive of post-target profitability. The predictor set is built using variables from the liquidity, leverage, asset structure, efficiency, reserves, size and growth of the company, excluding variables directly derived from profit. The explanatory feature space excludes direct profit-derived variables like net profit, profit before tax, earnings per share, and net profit margin. The final modelling sample for listed firms in India comprises 99 companies and 297 firm-year observations, based on financial statement data. The models are trained on observations through 2023 and tested on the held-out 2024 data. Eight models are compared: OLS, Ridge regression, firm fixed-effects OLS, Gradient Boosting, XGBoost, LightGBM, Support Vector Regression, and Multilayer Perceptron. Gradient Boosting performs best for ROE prediction, with holdout R² = 0.454, adjusted R² = 0.267, RMSE = 0.125, and MAE = 0.083. For ROA prediction, XGBoost outperforms the other models in terms of machine learning performance, with the following holdout performance metrics: R² = 0.470, adjusted R² = 0.288, RMSE = 0.055, and MAE = 0.041, whereas firm fixed effects OLS is competitive in ROA prediction. Lagged asset turnover, receivables turnover, asset growth, debt-to-equity, and book value per share are highlighted as key predictors, emphasising asset efficiency, working-capital conversion, growth dynamics, capital-structure effects, and balance-sheet strength. The sectoral projections for the 2026–2030 period are presented as bounded conditional scenarios rather than forecasts. The results demonstrate that by combining financial theory, econometric comparison, and transparent forecast assumptions, interpretable machine learning can assist profitability screening and sector benchmarking.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Predicting firm profitability in post-COVID India: A machine learning approach using fundamental financial ratios

  • 期刊论文

  • 发表日期：

    2026年7月30日

查看全文
  • Vijay Agrawal

摘要

This study proposes an interpretable machine learning model to predict a firm’s profitability in the context of post-COVID India, based on the previous year’s financial ratios. The dependent variables include Return on Equity (ROE) and Return on Assets (ROA), which are modelled one at a time to investigate whether pre-target financial fundamentals are predictive of post-target profitability. The predictor set is built using variables from the liquidity, leverage, asset structure, efficiency, reserves, size and growth of the company, excluding variables directly derived from profit. The explanatory feature space excludes direct profit-derived variables like net profit, profit before tax, earnings per share, and net profit margin. The final modelling sample for listed firms in India comprises 99 companies and 297 firm-year observations, based on financial statement data. The models are trained on observations through 2023 and tested on the held-out 2024 data. Eight models are compared: OLS, Ridge regression, firm fixed-effects OLS, Gradient Boosting, XGBoost, LightGBM, Support Vector Regression, and Multilayer Perceptron. Gradient Boosting performs best for ROE prediction, with holdout R² = 0.454, adjusted R² = 0.267, RMSE = 0.125, and MAE = 0.083. For ROA prediction, XGBoost outperforms the other models in terms of machine learning performance, with the following holdout performance metrics: R² = 0.470, adjusted R² = 0.288, RMSE = 0.055, and MAE = 0.041, whereas firm fixed effects OLS is competitive in ROA prediction. Lagged asset turnover, receivables turnover, asset growth, debt-to-equity, and book value per share are highlighted as key predictors, emphasising asset efficiency, working-capital conversion, growth dynamics, capital-structure effects, and balance-sheet strength. The sectoral projections for the 2026–2030 period are presented as bounded conditional scenarios rather than forecasts. The results demonstrate that by combining financial theory, econometric comparison, and transparent forecast assumptions, interpretable machine learning can assist profitability screening and sector benchmarking.