A Novel Hybrid Approach to Diagnosis of Parkinson's via Machine Learning Algorithms
Abstract – Parkinson's disease (PD) is a progressive, neurodegenerative disorder of aging that affects both motor and cognitive function. The etiology of PD is mostly unknown which makes the symptoms of the disease to be not easily identified. As a result, the objective of this investigation was to develop a machine-learning algorithm that diagnoses Parkinson's disease. The data used in this research were gathered from algorithms downloaded from uci.edu, and various algorithms including the Decision Tree, Logistic Regression, Random Forest, Adaptive Boosting, LGBM, and KNN algorithms were utilized. Our proposed model was combining the learning algorithms of K-means with another supervised learning algorithm. The symptoms analyzed are spread1, PPE, and MDVP.Fo(Hz), spread2, MDVP.Flo(Hz), MDVP.RAP and more which showed a result that the proposed model created the highest accuracy of 95.5% with the algorithm of fine-tuned LGBM. When we compare our product with other products of deep learning-based algorithms, it is clear that our product can be seen as more effective. First of all, there will be no need for our product to use GPU which will make the product much more cost-effective. Also, with these results, the proposed algorithm can be used for the development of a Parkinson's diagnosis kit that people can use during their daily lives to diagnose themselves.