An Effective Feature Selection Technique for Detecting Parkinson's Disease Using Binary Whale Optimization Algorithm

Parkinson's disease is a neurological condition that affects the nerves of the body and results in uncontrollable body movements like shaking, stiffness etc. Parkinson's disease affects around 10 million people worldwide. Early detection of the disease is very important for patients. The easiest way to detect the disease is with the help of the voice recordings i.e., speech signals. As the speech dataset has very large number of features it is important to choose the optimal feature subset for disease classification. This task is known as feature selection. This work proposes a feature selection approach for efficient detection of the Parkinson's disease based on binary whale optimization algorithm. Eight different transfer functions have been tried and the best one is chosen. The result has been evaluated based on classification accuracy, feature selection ratio and the execution time metrics. A significant improvement of 7.7 % have been achieved over the classical method.