Optimalisasi Klasifikasi Status Gizi Menggunakan Artificial Neural Network Dengan Pendekatan Explainable Artificial Intelligence Berbasis SHapley Additive Explanation (SHAP)

Optimization of Nutritional Status Classification Using Artificial Neural Networks with an Artificial Intelligence Explanation Approach Based on SHapley Additive Explanation (SHAP)

Kemajuan teknologi membutuhkan strategi yang tepat untuk menghasilkan analisis yang akurat dan optimal, khususnya di bidang kesehatan. Pembelajaran mesin (machine learning) menjadi salah satu pendekatan yang banyak diterapkan untuk membantu proses klasifikasi dan prediksi berbasis data medis. Salah satu permasalahan kesehatan yang memerlukan perhatian khusus adalah status gizi pada balita, baik kekurangan gizi maupun obesitas, yang berdampak jangka panjang terhadap pertumbuhan dan kualitas hidup anak. Penelitian ini dilakukan dalam dua tahap yang saling berkesinambungan. Tahap pertama berupa systematic literature review untuk mengidentifikasi algoritma machine learning yang paling efektif dan efisien dalam klasifikasi status gizi berdasarkan karakteristik dataset, metode yang digunakan, dan capaian akurasi. Hasil kajian menunjukkan bahwa Random Forest, Decision Tree, dan Naïve Bayes efektif untuk dataset kecil, XGBoost sesuai untuk dataset besar, serta kombinasi model memberikan performa optimal pada dataset menengah. Berdasarkan temuan tersebut, tahap kedua penelitian mengembangkan model eksperimen menggunakan Artificial Neural Network (ANN) yang dioptimalkan melalui pengaturan parameter pelatihan. Untuk meningkatkan transparansi model, diterapkan pendekatan Explainable Artificial Intelligence berbasis Shapley Additive Explanations (SHAP) guna menginterpretasikan kontribusi setiap fitur terhadap hasil prediksi. Evaluasi menggunakan metrik accuracy, precision, recall, dan F1-score menunjukkan bahwa model ANN yang dioptimalkan mampu memberikan performa klasifikasi yang baik sekaligus interpretasi yang jelas. Penelitian ini diharapkan dapat mendukung pengembangan sistem pendukung keputusan yang akurat, efisien, dan mudah dipahami dalam pemantauan status gizi balita.

Kata Kunci: Jaringan Syaraf Tiruan; Pembelajaran Mesin; SHAP; Explainable AI; Klasifikasi Obesitas.

Technological advancements require precise strategies to produce accurate and optimal analysis, particularly in the health sector. Machine learning is one approach widely applied to assist classification and prediction processes based on medical data. One health issue that requires special attention is the nutritional status of toddlers, both malnutrition and obesity, which have long-term impacts on children's growth and quality of life. This research was conducted in two interconnected stages. The first stage was a systematic literature review to identify the most effective and efficient machine learning algorithms for nutritional status classification based on dataset characteristics, methods used, and accuracy achieved. The results of the study showed that Random Forest, Decision Tree, and Naïve Bayes were effective for small datasets, XGBoost was suitable for large datasets, and a combination of models provided optimal performance for medium-sized datasets. Based on these findings, the second stage of the research developed an experimental model using an Artificial Neural Network (ANN) optimized by adjusting training parameters. To increase model transparency, an Explainable Artificial Intelligence approach based on Shapley Additive Explanations (SHAP) was applied to interpret the contribution of each feature to the prediction results. Evaluation using accuracy, precision, recall, and F1-score metrics demonstrated that the optimized ANN model delivered good classification performance and clear interpretation. This research is expected to support the development of accurate, efficient, and easy-to-understand decision support systems for monitoring the nutritional status of toddlers.

Keywords: Artificial Neural Network; Machine Learning; SHAP; Explainable AI; Classification of Obesity.