Analisis Komparasi Algoritma XGBoost dan Logistic Regression Berbasis Explainable AI (SHAP) untuk Deteksi Hipertensi
DOI:
https://doi.org/10.30865/jurikom.v13i3.9748Keywords:
Hypertension, Machine Learning, XGBoost, Explainable AI, SHAPAbstract
Hypertension represents a significant global health challenge due to its frequently asymptomatic nature until serious complications arise. This study aims to conduct an in-depth comparison between XGBoost and Logistic Regression algorithms for the early detection of hypertension, with a specific focus on explainability utilizing the SHAP Permutation Explainer method. The primary issue addressed is the tendency of previous research to manipulate datasets through oversampling techniques, which risks altering the natural distribution of clinical features, alongside the presence of "explainer bias" resulting from the use of non-uniform explanation methods. This research utilizes original data from the 2015 Behavioral Risk Factor Surveillance System (BRFSS) to ensure clinically reliable results. The results demonstrate that the XGBoost model consistently outperforms Logistic Regression across all evaluation metrics, with XGBoost achieving an AUC of 0.805 and an accuracy of 73.4%, compared to Logistic Regression's AUC of 0.802 and accuracy of 73.1%. SHAP-based interpretation reveals that BMI and Age are the most dominant predictors in both models. The primary contribution of this study is the provision of a predictive framework leveraging the original data distribution, yielding more objective predictions compared to synthetic manipulation, thereby serving as a credible reference for medical practitioners to understand hypertension risk factors in a transparent and accountable manner.
References
[1] M. Puji Lestari and A. M. Rusydan, “Penyuluhan Hipertensi dan DAGUSIBU Obat Antihipertensi,” J. Innov. Community Empower., vol. 6, no. 2, pp. 79–85, Sep. 2024, doi: 10.30989/jice.v6i2.1400.
[2] R. M. Touyz, “Hypertension 2022 Update: Focusing on the Future,” Hypertension, vol. 79, no. 8, pp. 1559–1562, Aug. 2022, doi: 10.1161/HYPERTENSIONAHA.122.19564.
[3] Y. A. Mashuri, N. Ng, and A. Santosa, “Socioeconomic disparities in the burden of hypertension among Indonesian adults - a multilevel analysis,” Global Health Action, vol. 15, no. 1, Dec. 2022, doi: 10.1080/16549716.2022.2129131.
[4] E. Martinez-Ríos, L. Montesinos, M. Alfaro-Ponce, and L. Pecchia, “A review of machine learning in hypertension detection and blood pressure estimation based on clinical and physiological data,” Biomedical Signal Processing and Control, vol. 68, p. 102813, Jul. 2021, doi: 10.1016/j.bspc.2021.102813.
[5] L. A. AlKaabi, L. S. Ahmed, M. F. Al Attiyah, and M. E. Abdel-Rahman, “Predicting hypertension using machine learning: Findings from Qatar Biobank Study,” PLoS ONE, vol. 15, no. 10, p. e0240370, Oct. 2020, doi: 10.1371/journal.pone.0240370.
[6] N. Chen et al., “Evaluating the risk of hypertension in residents in primary care in Shanghai, China with machine learning algorithms,” Front. Public Health, vol. 10, p. 984621, Oct. 2022, doi: 10.3389/fpubh.2022.984621.
[7] S. Montagna et al., “Machine Learning in Hypertension Detection: A Study on World Hypertension Day Data,” J Med Syst, vol. 47, no. 1, Dec. 2022, doi: 10.1007/s10916-022-01900-5.
[8] F. Si, Q. Liu, and J. Yu, “A prediction study on the occurrence risk of heart disease in older hypertensive patients based on machine learning,” BMC Geriatr, vol. 25, no. 1, p. 27, Jan. 2025, doi: 10.1186/s12877-025-05679-1.
[9] E. Bisong, N. Jibril, P. Premnath, E. Buligwa, G. Oboh, and A. Chukwuma, “Predicting high blood pressure using machine learning models in low- and middle-income countries,” BMC Med Inform Decis Mak, vol. 24, no. 1, p. 234, Aug. 2024, doi: 10.1186/s12911-024-02634-9.
[10] S. Widodo, H. Brawijaya, and S. Samudi, “Stratified K-fold cross validation optimization on machine learning for prediction,” SinkrOn, vol. 7, no. 4, pp. 2407–2414, Oct. 2022, doi: 10.33395/sinkron.v7i4.11792.
