Around 60 years ago, it was established that Cerebral Blood Flow (CBF) responds within fixed-time windows to changes in the systemic Arterial Blood Pressure (ABP). This finding and better technology motivated much research in the 1990s, giving rise to a new research field named Cerebral Hemodynamics (CH), which analyzed the transient states of these two and other signals in humans. Nowadays, it is widely accepted that at least three different mechanisms actuate on the control of CBF. One is the mechanism responsible for maintaining a constant CBF within a fairlywide range of ABP, which is referred to as the cerebral autoregulation. Another one is related to a metabolic regulation for keeping the balance between the local demand and the local supply of energy and oxygen in the brain, which is termed neurovascular coupling. The third mechanism corresponds to the so-called cerebrovascular reactivity, which regulates the vascular responses according to the concentration of certain gases in the blood. Currently, there is mounting evidence that CH is altered for some cerebrovascular pathologies, such as subarachnoid hemorrhage, severe head injury, vascular dementia and Alzheimer’s disease. But the evidence is not so clear for other disorders such as stroke and Parkinson’s disease, as example of problems that have shown a sharp increase in their prevalence in last few years. Although there exists a wide variety of methods to model the underlying mechanisms of CH, most of them are linear approximations. Moreover, they are normally used in combination with linear assessment tools of the quality of CBF control. The few nonlinear initiatives to model CH have not been adequately evaluated on data from patients with a cerebrovascular disorder, least of all with neurodegenerative diseases such as Parkinson’s. In this work, the modeling of CH with nonlinear machine learning methods is proposed, as they have proved their superiority in term of predictive accuracy. These models will be used jointly with a general assessment tool that is not based on linear structures. This combination could help to improve the current understanding of CH and might constitute a valid tool for the diagnosis of neurodegenerative disorders.