Chandroday Prakash Tiwari, C.N. Ram, Indivar Prasad, Rajesh Kumar, Aastik Jha and Indresh Kumar Tiwari
Abstract
High temperature is a major abiotic stress affecting growth and yield in chilli (Capsicum annuum L.). Therefore, the present investigation was undertaken to classify chilli genotypes for heat stress tolerance using machine learning approaches integrated with morpho-physiological and biochemical traits. Thirty-five chilli genotypes were evaluated under high temperature conditions during the SpringâSummer season of 2025 at ICAR-IIVR, Varanasi. Observations were recorded on important stress responsive traits including plant height, leaf area ratio, canopy temperature depression, stomatal density, antioxidant activity and viable pollen percentage. Correlation and Principal Component analyses revealed substantial variability and interrelationship among the studied traits, with the first five principal components explaining 77.07 per cent of the total variation. K-means clustering analyses effectively classified the genotypes into distinct groups. Random Forest (RF) and Support Vector Machine (SVM) models exhibited high classification efficiencies with accuracies of 97.14 and 96.67 per cent, respectively. Random Forest analysis identified leaf area ratio, plant height, catalase activity, sugar content and stomatal density as major contributing traits associated with heat stress tolerance. Genotypes G4, G6, G8, G10, G11, G12, G16, G21, G25, G29, G33 and G34 were identified as comparatively heat tolerant. The study demonstrated that machine learning integrated with physiological and biochemical traits provides an efficient approach for rapid identification of heat-tolerant chilli genotypes.