Laxmi Gulappagol, Pushparani M K, Dr. Harish Rajurkar, Prof. S. Pratap, Mohd Nawazuddin, Richa Malaviya · Computer and decision making. 2026 · 2026
DOI: 10.1922/comdem.v3i6s.18836
Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).
In high-stakes applications, such as those for intelligent decision support, the importance of explainable artificial intelligence (XAI) is growing as users need to understand, validate and act accordingly on the outputs of models. In this study, an explainable intelligent decision support workflow will be designed and tested with the case study of UCI Cleveland heart disease dataset. The number of observations were 303 and the number of predictors 13 with leakage-controlled preprocessing, stratified train-test separation, training only cross validation and comparison with logistic regression, decision tree, random forest and gradient boosting models. Random forest had the highest mean cross-validated ROC-AUC (0.904) and was therefore a priori chosen as the main model. On the held-out test set, it achieved ROC-AUC 0.947, sensitivity 0.929, specificity 0.818, F1 0.867, and balanced accuracy 0.873. Permutation importance demonstrated that the most significant predictors were number of major vessels, type of chest pain, thalassemic status, exercise induced angina and ST depression. Using the probability bands, model outputs were also converted into decision-support categories that make sense to people. These results reveal the potential of combining rigorous validation with explainability and risk stratification to boost intelligent decision support and indicate areas of sample size, external validation and clinical threshold generalizability.
No comments yet — start the discussion below.