Machine learning modeling for prediction of sweet potato slices quality based on colour property

Processing has effects on some key quality parameters either positively or negatively, there is also need for process monitoring to determine the state of processing. This study was set out to model the colour changes during the drying of Sweet Potato Slices as one of the key parameters of quality. The drying was carried out using two drying methods- hot air drying, HA and Vacuum drying VD. The sweet potato was dried at 3 mm thickness and isothermally at 70, 60 and 50 °C for both drying methods. Samples were blanched at 80 °C in hot water prior to drying. The color were measured using hunters method- a*, b*, L* and ∆E. The color were classified as under-dried, properly-dried, and over-dried and the machine learning algorithms used were namely Random Forest (RF), K-Nearest Neighbor (KNN), and Support Vector Machine (SVM) to trained on the dataset. The results indicated a superlative prediction accuracy of 100% by the RF model, followed by KNN with prediction accuracy of 98.3% and SVM with prediction accuracy of 95.7%. The research identified random forest model as the model with the best predictive capability among others, and therefore recommends its use in monitoring of the drying of potato slices given the conditions of this study.