Accurate prediction of avocado ripeness using image processing and machine learning models (multi-layer perceptron, support vector regression, and K-nearest neighbour)

Abstract Avocado is a popular fruit among health-conscious consumers, but it cannot be eaten raw due to tannins, which cause bitterness. Ripeness is commonly assessed by external characteristics, dry weight, fat content, and firmness. This study investigates the non-destructive prediction of ripeness in Hass avocado (Persea americana cv. “Hass”) using image processing combined with machine learning models, including multi-layer perceptron (MLP), support vector regression (SVR), and K-nearest neighbours (KNN), Skin colour features were extracted from the CIELab colour space (L*, a*, and b*) and used as input variables together with firmness measurements. The dataset was divided into 70/30, 80/20, and 90/10 ratios for training and testing. The results demonstrate that all three models are capable of predicting avocado firmness; however, the MLP model outperformed SVR and KNN. In particular, the MLP model with a 90/10 data split achieved the highest predictive performance, with an R2 of 0.990, a root mean square error of 0.341, and an MAE of 0.321. These findings indicate that the integration of image processing and machine learning provides an effective and reliable approach for predicting Hass avocado ripeness. The proposed method shows strong potential for practical applications, including the development of smartphone-based tools for real-time avocado ripeness assessment.