ABSTRACT
Parkinson's disease is a nervous system abnormality marked by decreased dopamine levels in the brain. Parkinson's disease inhibits one's ability to move. Speech difficulty, changes in movement and handwriting, and other symptoms are common with Parkinson's disease. A collection of hand drawings is employed to predict Parkinson's disease. There are 102 spiral images in the hand drawing dataset. Due to the minimal size of the dataset, augmentation is utilized to increase it. After that, the augmented images are utilized to train several machine learning and deep learning models, as well as pre-trained networks like RESNET50, VGG16, AlexNet, and VGG19. The performance metrics of hybrid models of deep learning with machine learning and hybrid models of deep learning (for feature extraction) with deep learning (for classification) are then compared. It was observed that the hybrid model of RESNET-50 and SVM performed well with better performance measures compared to other Machine Learning, Deep Learning and Hybrid Models with an accuracy score of 98.45%, sensitivity score of 0.99 and specificity score of 0.98.