Artificial intelligence (AI)-aided predictive modelling for yield and shelf-life enhancement in sweet pepper ( Capsicum annuum var. grossum )

Improving yield and postharvest shelf life is a key objective in sweet pepper breeding. The present study integrated artificial intelligence (AI) models to identify the major characters influencing yield and shelf life in sweet pepper grown under a side-ventilated, low-cost polyhouse. Twelve diverse genotypes were evaluated during 2021–2023 for morphological, quantitative, and quality traits. Quantitative characters showed high broad-sense heritability. The number of fruits per plant and fruit weight were positively correlated with yield per plant. Principal component analysis associated PC1 with yield-attributing traits, while PC2 was driven by shelf-life and quality-related characters. Multiple linear regression (MLR) revealed that the number of fruits per plant, stigma length, and fruit weight were key predictors of yields. However, the model does not converge for shelf life. A decision tree model identified fruit weight, total sugar, and total soluble solids as key determinants of shelf life. The multilayer perceptron (MLP) model further highlights leaf color, fruit color, and fruit orientation as important predictors of both yield and shelf life. Overall, the study provides AI-driven selection guidelines for identifying high-yielding, shelf-stable sweet pepper genotypes for protected cultivation. Future validation across larger populations will enable genomic-integrated precision breeding for sweet pepper improvement.