Deep Feature Selection from the Vocal Features for Effective Classification of Parkinson ’s disease
Parkinson’s disease (PD) is a neurological illness, which mostly impacts the movement of the body parts. We can see the severe impact of PD on individuals who are on or above 60 years old. PD can be easily acquired by men more than women. Parkinson’s disease initially seedbed in the brain, later it affects the movement of the body parts. PD affects the nerve cells called SubstantiaNigra(SN). SN is a comparatively small but a vital part of a human brain, which helps in the regulation of motor movements in a human anatomy. PD is gradationally increases, so it is a challenge for the physicians to discover early symptoms. It becomes worsens and leads to death when we couldn’t diagnose the PD at the early stages. The vital prognostics of PD are Shivering, Loss of Stiffness, Constipation, Tremor, and Vocal disorders. Early detection of PD, can improve the condition of a person via proper treatment but early diagnose is the key challenge. For this work, the various modulations of voice recordings of patients diagnosed with PD and healthy individuals were taken to train the feature subset to learn the algorithm. Initially, the millions of raw data have been pre-processed with several steps like Segmentation, Eliminating missing values and outliers, Dimensionality Reduction, etc. The dataset employed for this research work is obtained from University of California, Irvine (UCI) - a dedicated data library for Machine Learning (ML).The Correlation Feature selection algorithm (CFS) is a filter-based feature selection algorithm which will separate the features that are highly correlated with the classifier. In addition to it, Ant Colony Optimization (ACO), a Metaheuristic optimization strategy is incorporated to increase the preciseness of the data by optimizing the features selected by CFS and the optimized data will be gone through four supervised machine learning classifiers to get the final results. The final result reveals that the mixture of the proposed system gives the best results when compared to other combinations.