Feature Driven Machine Learning Models for Early Parkinson's Disease Detection in Healthcare Datasets

The nervous system of a person serves as the control centre for all physical activities. It causes thoughts to arise in the mind, produces responses, and makes it possible for body parts to move. It further controls additional functions, including digestion and breathing. This nervous system is made up of billions of cells that can translate the information received by sensory organs into actions carried out by body parts. However, as we get older, these cells begin to die, which causes the amount of the dopamine compound to decrease. This causes unexpected progression in the movement of the body parts, which is known as Parkinson's disease (PD). Although the disease is very challenging to cure but its production can be decelerated with early prediction. In this work, five machine learning models namely, random forest, decision tree, linear regression, naive bayes, and logistic regression have been used for the early PD diagnosis. For the implementation of the model, the dataset has been collected from Kaggle containing the normal cases and PD patients. The results have depicted that random forest models outperform with an accuracy rate of 86.24%, however, the naive shows the least accuracy value of 73.94%. Furthermore, the work can be helpful in diagnosing the active cases of PD in real-time medical treatment if enabled by the healthcare departments. In future, the work can be extended with the enhanced dataset and deep learning models can be implemented for better prediction outcomes.