Performance of Advanced Machine Learning Models in the Prediction of Amylose Content in Rice Using Internet of Things-Based Colorimetric Sensor

Rice ageing is a complicated process that is difficult to examine methodically. Several physicochemical properties of rice change with age as a function of moisture content and storage temperature. Among these qualities, amylose content is the most important and numerous metrics depend on it. Several sensors, Internet of Things, Information and Communication Technology, artificial intelligence and machine learning (ML) approaches are being used in technological interventions to tackle this problem. In the present study, seven advanced ML models were evaluated to classify the different concentrations of amylose using light-intensity data obtained by the novel colorimetric amylose sensor. From the performance of the evaluated ML models, it was observed that for the light intensity dataset obtained from the sensor, higher and similar model parameters and an accuracy value of 0.77 were observed for both artificial neural network (ANN) and k-nearest neighbour (KNN) algorithms, followed by accuracy values of 0.75, 0.74, 0.65, 0.61 and 0.61 respectively, for the decision tree, random forest, AdaBoost, logistic regression and support vector machine algorithms. Thus ANN and KNN are promising in predicting the different classes of amylose in rice.