The pressing need for effective follow-up biomarkers in ischemic stroke (IS) patients during the chronic phase finds a promising solution in machine learning (ML) techniques. Our study addresses this urgency by exploring noninvasive, accessible, and cost-effective tools to bridge the need in the primary care stroke gap. By leveraging 24-hour electrocardiography as an electrodiagnostic method for investigation the etiology of IS, we obtain Heart Rate Variability (HRV) parameters throughout the sleep-wake cycle. Our approach employs the k-fold cross-validation method on five ML models: random forest (RF), decision tree (DT), support vector machine (SVM), multilayer perceptron (MLP), and logistic regression (LR), aiming to pinpoint the optimal model for IS detection on both clinical variables and HRV parameters. Our results demonstrate that the RF performs best in detecting IS patients with remarkable accuracy, sensitivity, and specificity. Notably, our relevance analysis revealed the pivotal role of autonomic balance features, including time-domain long-term measures and vagal activity-related features, in influencing model performance. In this context, RF emerged not only as an IS detection model but also as a promising follow-up autonomic biomarker tool. This research highlights the need for personalized and efficient care in the management of ischemic stroke patients during the chronic phase, promoting a strategy for identifying IS.