Klasifikasi Bahan Pangan berdasarkan Kandungan Zat Gizi Bahan Pangan Menggunakan Fuzzy K-Nearest Neighbor

Abstract

The concept of glycemic index is an approach to select the right food for diabetics. One obstacle to implement this concept is the limited amount of known glycemic index data. With this knowledge about food nutrient and cooking process as factors that influence the glycemic index, we proposed a model to classify food based on its nutrients using fuzzy k-nearest neighbor algorithm. The results revealed that the classifier performance was not optimal because of the imbalanced data. The best accuracy for this model was only 53.69%. To overcome imbalanced data condition, we applied 4 different resampling techiques. The results showed that resampling techniques successfully solved the imbalanced data problem. In general, random over-sampling technique had the best performance with an accuracy of 84.39% in 1-NN. The next technique that had the best performance was synthetic minority over-sampling technique (SMOTE) with an accuracy of 73.37% in 1-NN and 2-NN.