Implementation of C4.5 Algorithm for Classification of Nutritional Status of Toddlers

Abstract

Nutrition is an important part in the growth of toddlers. Monitoring of nutritional status is needed. The data mining method used to classify the nutritional status of toddlers using C4.5 algorithm. The nutritional status of toddlers is divided into five classes, namely undernutrition, good nutrition, risk of overnutrition, overnutrition and obesity. There is an imbalance of data in the five classes. This data imbalance is handled using Synthetic Minority Oversampling Technique (SMOTE). From the research that has been conducted, the application of SMOTE in the classification of nutritional status of toddlers can influence the value of the model evaluation. Before SMOTE was applied, the classification model produced 86% accuracy, 87% precision, 86% recall, 85% f1-score, and 33% mean absolute error. After implementing SMOTE, it can increase the accuracy value to 90%, precision to 91%, recall to 90%, f1-score 90%, and can reduce the mean absolute error value to 22%.