Preliminary Evaluation of Support Vector Machine Models for Predicting Pear Fruit Quality During Shelf Life

This study explores predictive modeling of the physical, biochemical, and decay‐related properties of Spadana pears under varying shelf‐life conditions using machine learning. Pear samples were analyzed for moisture, biochemical, and physical attributes, with nondestructive CT imaging tracking structural changes. Support vector machines (SVMs) with different kernels predicted quality indicators, significantly impacting shelf life conditions. Gaussian and sigmoid kernels achieved high accuracy ( R 2 > 0.9) for firmness and volume, whereas linear kernels excelled in biochemical properties like phenol content ( R 2 = 0.95). Decay indicators showed moderate predictability, with polynomial and sigmoid kernels effectively capturing nonlinear interactions. This study represents a preliminary investigation based on a limited number of independent experimental units and a single pear cultivar under controlled laboratory conditions. Due to the exploratory nature of the work, the findings should be interpreted as indicative rather than conclusive and further validated in broader, real‐world postharvest scenarios.