AI-ENABLED PREDICTIVE MODELING OF CROP QUALITY AND POST-HARVEST FOOD LOSS INTEGRATING AGRICULTURAL AND FOOD-QUALITY INDICATORS

Abstract

This study created and tested AI-based predictive models for food loss in the post-harvest chain, while combining agricultural, storage, handling and food quality indicators. A structured data set was used, with 1,200 observations and 29 variables for six crops and five geographical regions, where the main outcomes are crop-quality score, crop-quality class, post-harvest loss percentage and loss-risk class. Climatic conditions, soil moisture, nutrient inputs, harvest maturity, storage and transport characteristics, soluble solids, pH, firmness, color index and defect rate were all used as agricultural, post-harvest and food-quality predictors. Linear Regression, Decision Tree, Random Forest, XGBoost, Support Vector Machine, and K-Nearest Neighbors were tested using regression and classification metrics. It was revealed that agricultural condition score, colour index, harvest maturity and defect rate were significant variables influencing the quality of the produce while the storage temperature, soluble solids, pH, handling score, storage duration and defect rate were significant for post-harvest loss. Logistic Regression and SVM gave better-balanced classification results with R² values of 0.894 and 0.874 respectively, while Linear Regression gave R² values of 0.725 and 0.820 for crop quality and post-harvest loss respectively. The results highlight the usefulness of combining information related to agricultural production and food quality for prediction and post-harvest management based on data.