Prediction of Parkinson’s Disease Using Machine Learning Dungala Joshna

Abstract

Parkinson’s Disease (PD) is a disorder that neurodegenerative in nature in addition to affecting the motor and speech functions of an individual. As a result,
proper treatment as well as management becomes easy with the earliest diagnosis possible. Unlike the healthcare practitioners’ diagnosis which employs
machine learning techniques, the traditional methods lack the use of sophisticated technology and make use of expert diagnosis which is bound to take many
hours or days to complete. This work exhibits how deep learning can help in devising frameworks that predict the chance of Parkinson’s Disease using voice
recordings obtained from a publicly accessible dataset at UCI Machine Learning Repository that features other machine learning programs like Logistic
Regression, Support Vector SVM, Random Forest, and K-Nearest Neighbours. The voice recordings were subjected to primary screening methods like
normalization all the way to final ones like filter feature selection in order to pave the way for optimal performance of the models while undergoing training
along with standard validation tools such as accuracy, precision, recall along with F1 value and ROC AUC, as well. [Add in best performing models’ name
here] emerged as the most accurate of the models in prediction accuracy which confirms the preliminary capabilities of machine learning to help with timely
and precise diagnosis of Parkinson's Disease. The work goes a step further in not just showcasing the power of computational methodologies in medicine but
also serves as a platform towards devising real-time diagnostic applications.