Jurnal Sisfokom (Jul 2025)
The Effect of the SMOTE Method on the Classification of Toddler Nutritional Status Using the Naïve Bayes Method
Abstract
The first five years of life are a golden age for growth and development, so fulfilling nutritional intake during this period is very important to avoid stunting or growth failure. The problem of stunting is still the focus of the government because it is related to nutrition which is one of the key aspects for the development of qualified resources as well as in national development. According to the report of the Ministry of Health in 2023, it was stated that the results of the 2023 Indonesian Health Survey showed that there had been a decreasing in the prevalence of stunting over the past 10 years but it had not been able to meet the target of the 2020-2024 National Medium-Term Development Plan of 14% in 2024. This study will classify the toddler’s nutritional status using the Naive Bayes method. This method uses a probability technique with Bayes' theorem which is based on the assumption of mutually independent and equal conditions. The calculation of the Naive Bayes probability in this study uses the Multinomial distribution because the data used is discrete data. The total numbers of toddlers’ nutritional status data obtained was 245 data, with 4 invalid data. Based on the data set owned, the number of samples for each class label had an unbalanced number. One method could be used to handle this unbalanced data is the random oversampling method, Synthetic Minority Oversampling (SMOTE). SMOTE will create synthetic data randomly to balance minority data samples. The analysis and testing results showed that in Multinomial Naive Bayes with the 10-cross validation technique, the g-means value obtained on the original data set was 44.98% while in the balanced data set the g-means value was 80.06%. In Multinomial Naive Bayes with the split validation technique, the g-means value obtained on the original data set was 44.20% while in the balanced data set was 80.06%. This showed that there was an increase in the g-means value of 35%. It can be stated that the SMOTE method effectively improves the overall capability of the Multinomial Naive Bayes model.
Keywords
