Analisis Kontribusi Fitur Bertahap dan Stabilitas Penjelasan SHAP pada Prediksi Risiko Diabetes Menggunakan XGBoost
DOI:
https://doi.org/10.30865/jurikom.v15i5.10234Keywords:
Prediksi Diabetes, XGBoost; , XGBoost, SHAP, Stabilitas Penjelasan, Data Tidak SeimbangAbstract
Model prediksi diabetes berbasis data tabular kerap dilaporkan dengan akurasi di atas 96%, padahal rendahnya prevalensi kasus membuat akurasi tersebut condong pada kelas mayoritas. Selain itu, fitur non-invasif dan penanda diagnostik umumnya digabungkan dalam satu model, sementara penjelasan SHAP jarang diuji kestabilannya. Penelitian ini bertujuan mengukur kontribusi marginal setiap kelompok fitur sekaligus menguji stabilitas penjelasan model. Sebanyak 96.128 rekam data dengan prevalensi diabetes 8,82% dimodelkan dalam empat tahap fitur, yaitu demografi dan riwayat penyakit (T0), penambahan indeks massa tubuh (T1), glukosa darah (T2), dan HbA1c (T3). XGBoost yang dioptimasi dengan Optuna dibandingkan dengan ResMLP dan FT-Transformer melalui validasi silang berulang 5×3 serta data uji terpisah, dengan AUPRC, balanced accuracy, dan Brier score sebagai metrik utama. AUPRC XGBoost mencapai 0,266; 0,325; 0,713; dan 0,889 untuk T0 hingga T3, dan setiap kenaikan antartahap signifikan (uji t terkoreksi, p < 0,001). Glukosa darah memberikan kenaikan AUPRC terbesar (0,389), sedangkan kontribusi usia menyusut dari 58,9% menjadi 12,7%. Peringkat fitur SHAP tetap stabil (Kendall τ = 0,93–1,00; Jaccard tiga fitur teratas = 1,0) walaupun kinerja model berbeda jauh antartahap. XGBoost setara dengan model deep learning pada T0–T1, unggul pada T2–T3, dan lebih terkalibrasi. Audit menemukan 23,1% label tidak sesuai ambang ADA, dan seluruh kesalahan T3 berada pada zona label tidak konsisten. Fitur non-invasif berpotensi menjadi alat triase, sedangkan diagnosis membutuhkan penanda glikemik; temuan dari data publik berpola sintetis ini memerlukan validasi klinis eksternal.
References
[1] International Diabetes Federation, “IDF Diabetes Atlas, 11th Edition,” International Diabetes Federation, Brussels, 2025. [Online]. Available: https://diabetesatlas.org/resources/idf-diabetes-atlas-2025/
[2] H. Sun, P. Saeedi, S. Karuranga, M. Pinkepank, K. Ogurtsova, and B. B. Duncan, “IDF Diabetes Atlas: Global, regional and country-level diabetes prevalence estimates for 2021 and projections for 2045,” Diabetes Research and Clinical Practice, vol. 183, p. 109119, 2022, doi: 10.1016/j.diabres.2021.109119.
[3] Badan Kebijakan Pembangunan Kesehatan, “Survei Kesehatan Indonesia (SKI) 2023 dalam Angka,” Kementerian Kesehatan Republik Indonesia, Jakarta, 2023. [Online]. Available: https://www.badankebijakan.kemkes.go.id/hasil-ski-2023/
[4] M. Mustafa, “Diabetes Prediction Dataset,” Kaggle, 2023. [Online]. Available: https://www.kaggle.com/datasets/iammustafatz/diabetes-prediction-dataset
[5] N. A. N. A. Fauzi and F. Budiman, “Analisis Perbandingan Metode Random Forest, XGBoost, dan Logistic Regression Untuk Klasifikasi Deteksi Dini Penyakit Diabetes,” JURIKOM (Jurnal Riset Komputer), vol. 12, no. 6, pp. 995–1005, 2025, doi: 10.30865/jurikom.v12i6.9392.
[6] M. R. Khurshid, S. Manzoor, T. Sadiq, L. Hussain, M. S. Khan, and A. K. Dutta, “Unveiling diabetes onset: Optimized XGBoost with Bayesian optimization for enhanced prediction,” PLoS One, vol. 20, no. 1, p. e0310218, 2025, doi: 10.1371/journal.pone.0310218.
[7] W. Li, Y. Peng, and S. Xiong, “Diabetes Risk Prediction Model Based on Improved XGBoost with Genetic Algorithm,” in Proceedings of the 2024 4th International Conference on Computational Modeling, Simulation and Data Analysis, 2024, pp. 390–396. doi: 10.1145/3727993.3728059.
[8] A. B. Hidayah and A. N. Ikhsan, “Analisis Performa SMOTE dan SMOTE-TomekLink untuk Klasifikasi Diabetes: XGBoost-SHAP,” Journal of Software Engineering and Information System (SEIS), vol. 6, no. 2, pp. 84–94, 2026, doi: 10.37859/seis.v6i2.11357.
[9] W. Iftikhar, M. Yaseen, G. Rahman, M. A. Nauman, and U. F. Khattak, “Improving Diabetes Prediction Accuracy and Interpretability with SMOTE and SHAP,” Engineering, Technology & Applied Science Research, vol. 16, no. 2, pp. 34276–34282, 2026, doi: 10.48084/etasr.16247.
[10] D. T. Ramadhan and F. Agustina, “Prediksi Diabetes Mellitus dengan Ensemble Gradient Boosting dan Advanced Feature Engineering,” Building of Informatics, Technology and Science (BITS), vol. 7, no. 2, pp. 1222–1233, 2025, doi: 10.47065/bits.v7i2.8011.
[11] A. Nasrin, M. B. Asif, R. Naidu, A. H. Asif, M. S. Chohan, and G. A. Khan, “An Optimized and Explainable Machine Learning Framework for Diabetes Prediction Using Marine Predators Algorithm and SHAP,” Journal of Personalized Medicine, vol. 16, no. 8, p. 423, 2026, doi: 10.3390/jpm16080423.
[12] A. E.-S. El-Bashbishy and H. M. El-Bakry, “Pediatric diabetes prediction using machine learning,” Scientific Reports, vol. 16, no. 1, p. 1979, 2026, doi: 10.1038/s41598-025-24964-y.
[13] W. Li, Y. Peng, and K. Peng, “Diabetes prediction model based on GA-XGBoost and stacking ensemble algorithm,” PLoS One, vol. 19, no. 9, p. e0311222, 2024, doi: 10.1371/journal.pone.0311222.
[14] M. A. Taufiqurrahman and C. D. S. Pratama, “Klasifikasi Risiko Penyakit Metabolik dengan model Ensemble, Decision Tree Dan Logistic Regression Pada Posbindu Universitas Gadjah Mada,” JURIKOM (Jurnal Riset Komputer), vol. 13, no. 4, pp. 1230–1245, 2026, doi: 10.30865/jurikom.v13i4.9958.
[15] T. Saito and M. Rehmsmeier, “The Precision-Recall Plot Is More Informative than the ROC Plot When Evaluating Binary Classifiers on Imbalanced Datasets,” PLoS One, vol. 10, no. 3, p. e0118432, 2015, doi: 10.1371/journal.pone.0118432.
[16] N. V. Chawla, K. W. Bowyer, L. O. Hall, and W. P. Kegelmeyer, “SMOTE: Synthetic Minority Over-sampling Technique,” Journal of Artificial Intelligence Research, vol. 16, pp. 321–357, 2002, doi: 10.1613/jair.953.
[17] R. van den Goorbergh, M. van Smeden, D. Timmerman, and B. Van Calster, “The harm of class imbalance corrections for risk prediction models: illustration and simulation using logistic regression,” Journal of the American Medical Informatics Association, vol. 29, no. 9, pp. 1525–1534, 2022, doi: 10.1093/jamia/ocac093.
[18] American Diabetes Association Professional Practice Committee, “2. Diagnosis and Classification of Diabetes: Standards of Care in Diabetes—2025,” Diabetes Care, vol. 48, no. Supplement_1, pp. S27–S49, 2025, doi: 10.2337/dc25-S002.
[19] B. Sun, C. Zhang, K. Ji, and X. Lu, “Beyond the Blood Draw: Explainable Machine Learning for Non-Invasive Dysglycemia Risk Screening,” arXiv preprint arXiv:2606.16056, 2026, doi: 10.48550/arXiv.2606.16056.
[20] S. M. Lundberg and S.-I. Lee, “A Unified Approach to Interpreting Model Predictions,” in Advances in Neural Information Processing Systems 30 (NeurIPS 2017), 2017, pp. 4765–4774.
[21] J. Goldwasser and G. Hooker, “Statistical Significance of Feature Importance Rankings,” arXiv preprint arXiv:2401.15800, 2024, doi: 10.48550/arXiv.2401.15800.
[22] S. Krishna, T. Han, A. Gu, S. Wu, S. Jabbari, and H. Lakkaraju, “The Disagreement Problem in Explainable Machine Learning: A Practitioner’s Perspective,” Transactions on Machine Learning Research, 2024, doi: 10.48550/arXiv.2202.01602.
[23] P. Guillén and E. Frias-Martinez, “Enhancing SHAP Explainability for Diagnostic and Prognostic ML Models in Alzheimer’s Disease,” Computers, Materials & Continua, vol. 87, no. 2, p. 95, 2026, doi: 10.32604/cmc.2026.076400.
[24] Y. Gorishniy, I. Rubachev, V. Khrulkov, and A. Babenko, “Revisiting Deep Learning Models for Tabular Data,” in Advances in Neural Information Processing Systems 34 (NeurIPS 2021), 2021, pp. 18932–18943.
[25] L. Grinsztajn, E. Oyallon, and G. Varoquaux, “Why do tree-based models still outperform deep learning on typical tabular data?,” in Advances in Neural Information Processing Systems 35 (NeurIPS 2022) Datasets and Benchmarks Track, 2022, pp. 507–520.
[26] R. Shwartz-Ziv and A. Armon, “Tabular data: Deep learning is not all you need,” Information Fusion, vol. 81, pp. 84–90, 2022, doi: 10.1016/j.inffus.2021.11.011.
[27] T. Chen and C. Guestrin, “XGBoost: A Scalable Tree Boosting System,” in Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 2016, pp. 785–794. doi: 10.1145/2939672.2939785.
[28] T. Akiba, S. Sano, T. Yanase, T. Ohta, and M. Koyama, “Optuna: A Next-generation Hyperparameter Optimization Framework,” in Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, 2019, pp. 2623–2631. doi: 10.1145/3292500.3330701.
[29] A. Géron, Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow, 3rd ed. Sebastopol, CA: O’Reilly Media, 2022.
[30] H. Gultom, A. Fauzi, and R. Ester, “Optimasi Hyperparameter XGBoost Berbasis Bayesian Optimization dan SHAP untuk Prediksi Dropout Mahasiswa Perguruan Tinggi,” JURIKOM (Jurnal Riset Komputer), vol. 13, no. 4, pp. 1189–1200, 2026, doi: 10.30865/jurikom.v13i4.10055.
[31] M. Sundararajan, A. Taly, and Q. Yan, “Axiomatic Attribution for Deep Networks,” in Proceedings of the 34th International Conference on Machine Learning (ICML), 2017, pp. 3319–3328.
[32] C. Molnar, Interpretable Machine Learning: A Guide for Making Black Box Models Explainable, 3rd ed. Leanpub, 2025. [Online]. Available: https://christophm.github.io/interpretable-ml-book/
[33] C. Nadeau and Y. Bengio, “Inference for the Generalization Error,” Machine Learning, vol. 52, no. 3, pp. 239–281, 2003, doi: 10.1023/A:1024068626366.
