Parkinson's disease (PD) is a debilitating state that hinders quality of life. It results from the demise of dopamine-generating cells across the region of substantia nigra of the central nervous system (CNS), that has an impact on the body. Individuals with Parkinson's disease experience difficulty speaking, writing, and walking. To distinguish between individuals with Parkinson's disease and those in good health, various machine learning-based approaches are employed. A thorough assessment of methods based on machine learning for Parkinson disease prediction is presented in this work. With an emphasis on the application of research findings in real-world clinical settings, this review critically evaluates the usefulness of machine learning models regarding the practical identification of Parkinson’s disease. In this work, literature analysis was carried out on published papers in past years utilizing the PubMed, University of California, Irvine (UCI) repository, and Xplore databases on Institute of Electrical and Electronics Engineers (IEEE) offer an extensive summary among the techniques for machine learning and data modalities used in differential diagnosis of Parkinson's disease. We have proposed methodology also based on our study.