Cloud-Based Parkinson Disease Prediction System Using Expanded Cat Swarm Optimization

Parkinson disease is identified as the second most severe neurodegenerative disorder that affects the nerve system of people. This disease could mainly affect the walking, speech, and vision of patients followed by body nervousness, handwriting, harsh voice quality, depression, and sleeping problems. The proposed research study focuses on early detection and diagnosis of disease from the accelerometer sensor-based data by evaluating the deviations present in patient’s motor symptoms. A cloud-based Parkinson disease prediction system is developed for a clinical decision-making process that helps the doctor to diagnose the Parkinson-influenced patient from a remote place. Gait parameters of the patient were extracted along to provide input vectors to the classifier model for onboard Parkinson disease prediction and diagnosis. An effective expanded cat swarm optimization (ECSO)-based feature selection technique is explored to overcome the problem of dataset dimensionality. It could select the most appropriate features from the patient dataset according to a logically inspired evolutionary algorithm. Using this feature selection technique in the k-Nearest Neighbor (k-NN), classifier model could significantly improve the disease prediction accuracy, and also minimizes the disease prediction time against the existing classifier models.