DETECTION OF PARKINSON DISEASE USING MACHINE LEARNING
Abstract
The symptoms of Parkinson's disease (PD), the second most prevalent neurological disorder in the elderly, include a wide variety of impairments in motor control and cognitive development. Following a stroke, this is the most common neurological disorder. Because the symptoms are comparable to those of other diseases, Parkinson's disease (PD) may be hard to diagnose. This category includes disorders such as essential tremor and aging. About the time you hit fifty-five, you'll start to notice symptoms like trouble walking and speaking more often. Medication is available to alleviate some of the symptoms of Parkinson's disease (PD), but there is yet no cure. Assuming they can manage their symptoms, everyone can carry on with their regular lives. Recognizing this disease and intervening to prevent its progression is of the utmost importance. Attempts to determine the disease kind have necessitated extensive investigation. We are primarily focused on creating and applying various deep learning and machine learning models for the aim of diagnosing Parkinson's disease (PD). The Support Vector Machine (SVM), Random Forest (RF), Decision Tree (DT), Knowledge Network (KNN), and Multi-Layer Perceptron (MLP) are just a few examples of these types of models. Numerous models are at your disposal, not limited to KNN and MLP. The objective is to use the characteristics of the speech signals to differentiate between healthy individuals and those with Parkinson's disease (PD). The dataset, which included 150 audio recordings of exams given to 31 people, was retrieved from the UC Irvine machine learning repository. More than that, we improved our models' performance through training with features selection, hyperparameter tuning, and the Synthetic Minority Over-sampling Technique (SMOTE) (GridSearchCV). The most effective combination of GridSearchCV, SMOTE, MLP, and SVM, with a train/test split ratio of 70:30, yielded the greatest results for our project. To review, MLP attained a f1-score of 99%, an accuracy of 98.31%, a recall of 98%, and precision of 100%. Along with a remarkable 95% identification rate, 98% accuracy, 96% recall, and 97% f1-score, the support vector machine (SVM) also achieved 98% accuracy. It appears that the proposed method might be helpful for PD prediction and could be readily integrated into healthcare systems for diagnosis, according to the clinical trials conducted for this work.