Millions of people throughout the world suffer with Parkinson's disease (PD), severely reducing their quality of life. With the symptoms when we detect Parkinson disease automatically, it could provide insights to the disease's early stages of development, enhancing the patients' projected clinical results through correctly focused therapies. This potential has prompted numerous academics to explore ways for measuring and quantifying the existence of PD symptoms using commercially available sensors. In this paper, we offer an overview of some recent scientific articles on several machine learning techniques that assist physiologists in detecting PD early. In addition, a comparative study between traditional machine learning (TML) algorithms and deep learning (DL) architectures based on the scope of their appropriate usage for classifying PD effectively has been discussed. Based on the comparison on detecting the PD from previous works, this paper concludes that deep learning models are more efficacious than traditional machine learning algorithms.