Panarit Sakunasinha. Nondestructive measurement of dielectric property of Nam Dok Mai Si Thong mango fruit. Doctoral Degraee(Physics). Kasetsart University. Office of the University Library. : Kasetsart University, 2022.
Nondestructive measurement of dielectric property of Nam Dok Mai Si Thong mango fruit
Abstract:
This thesis presents an approach to classify the ripeness stage of mangoes based on physical, electrical, and biochemical attributes using four machine learning (ML) algorithms. First, the fruit attribute data were divided into two groups: one for training and another for testing the ML classifiers. Next, the K-means algorithm was used to analyze the biochemical data and to define the ripeness of mangoes into either two classes (unripe and ripe) or three classes (unripe, ripe, and overripe). The Silhouette and the Elbow methods were then employed to verify the optimum number of mango ripeness stages. Next, the synthetic minority oversampling technique was applied to balance the number of mango samples in each class. The classifiers were then provided with a group of the fruit attribute data for training, with the classbalanced data as the target. The performance of the ML models for ripeness classification was compared using four-fold cross validation. Finally, the best performing ML model was chosen for further testing against another mango attribute dataset. Results showed that the ML classifiers performed better with oversampled data than without. The feed-forward artificial neural network classifier performed the best classification, with a mean accuracy of 89.6% for three classes when compared to the other classifiers.
Kasetsart University. Office of the University Library