Andri Pranolo, Nurul Putrie Utami, Agung Bella Putra Utama, Fairuz Khairunnisa Anasyua, Itoshiko Nurahman, Aji Prasetya Wibawa
This study aims to evaluate and compare the performance of two obesity classification approaches using the Long Short-Term Memory (LSTM) method and a combination of Transfer Learning with LSTM (TL-LSTM). The research data consisted of demographic features, food consumption patterns, and physical activity that were used to classify obesity rates in several categories. The evaluation used the Accuracy, Precision, Recall, and F1 Score metrics and the confusion matrix. The results showed that the TL-LSTM model consistently outperformed the regular LSTM with an increase in accuracy from 85.82% to 89.60%, precision from 86.30% to 89.94%, recall from 85.82% to 89.60%, and F1 Score from 85.37% to 89.52%. In conclusion, the integration of transfer learning with LSTM has been proven to be able to significantly improve the performance of obesity classification compared to the use of LSTM independently. © 2025 IEEE.
Universitas Ahmad Dahlan, Informatics Department, Yogyakarta, Indonesia; Universitas Ahmad Dahlan, Department of Food Service Industry, Yogyakarta, Indonesia; Electrical Engineering and Informatics Universitas Negeri Malang, Malang, Indonesia; Universitas Ahmad Dahlan, Department of Nutrition, Yogyakarta, Indonesia
