The article examines equity price deviations in India by integrating behavioural, structural, and firm fundamentals within a Random Forest (RF) expected-price framework. Traditional valuation approaches such as discounted cash flow (DCF) and linear regression often struggle to capture nonlinear distortions, particularly in emerging markets where retail sentiment, social media activity, and structural factors play a decisive role. To address this gap, we construct a representative dataset covering 2005–2024 that combine firm-level fundamentals (EPS, ROE, EBITDA margin), structural attributes (promoter holdings, split history), behavioural indicators (Google Trends, sentiment scores), and market factors (volatility, FII/DII flows, and trading volume). The RF model is trained under an annual rolling-window design and optimized through grid search to ensure robustness across different market regimes. Predictive performance is evaluated using standard error metrics (RMSE, MAE, and directional accuracy), while interpretability is enhanced through SHAP values and scenario-based analysis. Price deviation is defined as the difference between the observed market price and the RF-estimated expected price benchmark, where positive values indicate relative overpricing and negative values indicate relative under-pricing against the model benchmark. The results show that firms with weaker fundamentals and elevated retail attention tend to exhibit larger expected-price deviations, while periods of heightened market stress are associated with more pronounced negative deviations among mid-cap stocks. Overall, the study proposes a transparent and replicable expected-price framework that integrates behavioural finance insights with machine learning, offering practical relevance for investors and regulators. It enriches equity price literature in emerging markets and underscores integrating non-linear drivers into valuation.
Natarajan et al. (Thu,) studied this question.