The selection of suitable cherry tomato (Solanum lycopersicum var. Cerasiforme) genotypes for Indian agro-climatic conditions is vital for maximizing profitability and achieving sustainable cultivation. Traditional genotype selection through field trials is time-consuming and resource-intensive. This study leverages machine learning (ML) models to classify cherry tomato genotypes based on yield and quality traits under both open-field and polyhouse conditions. A comprehensive dataset, comprising morphological, physiological, and biochemical parameters such as plant height, days to flowering, fruit morphology, lycopene, β-carotene, total soluble solids (TSS), sugars, and acidity, was used for model training and evaluation.

Protected cultivation significantly enhanced fruit quality, with polyhouse-grown tomatoes exhibiting up to 51.88% higher TSS (4.2–7.9 °Brix) and superior lycopene content (1.07–7.48 mg/100 g) compared to open-field conditions. Five ML classifiers—Decision Tree, Random Forest, Support Vector Machine (SVM), K-Nearest Neighbours (KNN), and Neural Network—were evaluated using 80:20 train–test split to ensure external validation. The Neural Network model achieved the best performance with an accuracy of 75%, F1-score of 0.66, and ROC-AUC of 0.88. The Decision Tree model showed comparable accuracy (75%) but a lower ROC-AUC (0.77). Random Forest and SVM achieved 50% accuracy with ROC-AUC values of 0.77 and 0.16, respectively, while KNN performed poor (accuracy: 25%, ROC-AUC: 0.50).

These results highlight the potential of ML-based classification in enhancing the efficiency of genotype selection, minimizing dependency on exhaustive field evaluations, and promoting precision agriculture. The findings serve as a decision-support tool for breeders and cultivators aiming to optimize genotype deployment under diverse cultivation environments.