Background and Aim: Ischemic heart disease, marked by decreased blood flow to the heart muscle, is a global health concern with notable morbidity and mortality rates. Investigating novel strategies like machine learning for risk assessment, specifically among intensive care unit (ICU) admitted ischemic stroke patients, holds the potential for better outcomes. This study develops a machine learning framework to predict the 2-year risk of ischemic heart disease in ICUadmitted ischemic stroke patients, enhancing long-term prognostic accuracy. Methods: Our study encompassed a cohort of 2,068 ischemic stroke patients admitted to the ICU from the period 2001 to 2012.We applied both a holdout strategy and a 10-fold cross-validation method during model development. Stepwise logistic regression was used to select predictors. We adopted two machine learning models such as random forest and XGBoost model for our prediction. Results: Among the 2,068 patients, 446 had IHD during a 2-year ICU follow-up, while 1,622 did not. Baseline findings revealed that the majority of IHD patients were male (64%), with a median age of 72 years. Both XGBoost and random forest models exhibited the same discriminative power, boasting an AUROC of 93%. Notably, the top five variables in our model were platelet count, potassium levels, age, troponin T, and magnesium levels. Conclusions: The comparative analysis highlights the superior performance of the random forest model in terms of sensitivity, specificity and accuracy, underlining its potential clinical utility for identifying high-risk patients and guiding interventions to mitigate IHD risk.