Evidence record · UDINUS · SDGs
- Unit
- Rektorat UDINUS
- Period
- 2025
- Goals
- Record
- e337dc12-7a82-4976-b04d-236b71c01746
- Published
- 2026-10-30
Universitas Dian Nuswantoro conducted a study evaluating Explainable Artificial Intelligence methods for cardiovascular disease prediction. The activity involved UDINUS researchers Nuraini, Aisyah; Shidik, Guruh Fajar; Widhiyanti, Erna Amalia; Ramadhani, Talitha Olga; and Kusuma, Edi Jaya, who participated in examining the interpretability of machine learning models. The study addressed the need for accurate and interpretable prediction models in the context of cardiovascular diseases, which are identified in the source as a leading cause of death globally.
Using a public dataset, the activity tested several machine learning models for cardiovascular disease prediction. AdaBoost achieved the best reported performance, with an AUC of 91% and accuracy of 85%. The researchers analyzed feature importance using SHAP and LIME. The consistency of the explanations generated by these methods was validated using Spearman Rank and Kendall's Tau correlations.
The findings showed that SHAP and LIME identified similar key features. However, SHAP was more aligned with statistical patterns and demonstrated better reliability. SHAP also provided clearer global interpretability, whereas LIME produced more varied, instance-specific insights. The study demonstrated SHAP's stronger performance in delivering consistent and data-aligned explanations for cardiovascular disease prediction. The findings support the integration of interpretable models in healthcare to enhance clinical decision-making and trust in artificial intelligence systems. Through this activity, UDINUS contributed to the examination of transparent prediction methods in health-related artificial intelligence.