Prediksi Status Gizi pada Balita Menggunakan Metode Long ShortTerm Memory

Abstract

The high prevalence of malnutrition among toddlers in Pontianak City, adversely affecting growth, cognitive function, and future human capital. Despite monitoring efforts, traditional methods remain slow and less accurate. The objective is to develop an early prediction system for toddler nutritional status using a Long Short-Term Memory (LSTM) model based on anthropometric data (weight, height, age, gender) from the Pontianak City Health Office. The methodology comprises an applied experimental quantitative design with the following stages: data collection and preprocessing (cleaning, normalization, train-test split), LSTM model architecture development, model training, and performance evaluation using accuracy, precision, recall, and F1-score metrics. The dataset includes 1,182 samples of toddlers aged 0–60 months. Training results yielded 82% accuracy, 98% precision, 82% recall, and an F1-score of 89% on the training set. On the test set, the model achieved 79% accuracy, 99% precision, 80% recall, and an F1-score of 89%, with a weighted average F1-score of 0.83 and a macro average F1-score of 0.71, indicating solid overall performance but highlighting the need for improvement in underrepresented classes. The study’s implications involve enhancing the effectiveness of toddler nutritional monitoring, enabling faster interventions by healthcare workers, and contributing academically to the development of AI-based applications in public health.