Parkinson’s disease (PD) is the second most common neurodegenerative disease in the world. The number of cases has been growing exponentially in recent years, and it remains without a cure to date. Diagnosing the disease is still a challenge, and research indicates that many diagnoses are made incorrectly, showing a statistically irrelevant increase in accuracy over recent years. Faced with this problem, different researchers have been developing devices to collect inertial data rich in patient attributes, using sensors capable of capturing a person’s movements and identifying the intensity of PD motor symptoms. The objective of this work was to analyze data from the research by Machado et al. [1], in which data were collected from different patients with PD and neurologically healthy individuals, and to evaluate the use of a machine learning model to estimate the probability of a person having motor symptoms of PD. As a result of this work, a dimensional reduction of the data was carried out, and a Binary Logistic Regression model was estimated, obtaining an average accuracy rate of approximately 90% on data outside the training sample. The estimated model presented satisfactory performance statistics for identifying people with Parkinson’s, and it was concluded that the diagnosis of PD could benefit from models like this.