Hyper-parameter optimization of deep learning model for prediction of Parkinson’s disease

Neurodegenerative disorder such as Parkinson’s disease (PD) is among the severe health problems in our aging society. It is a neural disorder that affects people socially as well as economically. It occurs due to the failure of the brain’s dopamine-producing cells to produce enough dopamine to enable the motor movement of the body. This disease primarily affects vision, speech, movement problems, and excretion activity, followed by depression, nervousness, sleeping problems, and panic attacks. The onset of Parkinson’s disease is diagnosed with the help of speech disorders, which are the earliest symptoms of it. The essential goal of this paper is to build up a viable clinical decision-making system that helps the doctor in diagnosing the PD influenced patients. In this paper, a specific framework based on grid search optimization is proposed to develop an optimized deep learning Model to predict the early onset of Parkinson’s disease whereby multiple hyperparameters are to be set and tuned for evaluation of the deep learning model. The grid search optimization consists of three main stages, i.e., the optimization of the deep learning model topology, the hyperparameters, and its performance. An evaluation of the proposed approach is done on the speech samples of PD patients and healthy individuals. The results of the approach proposed are finally analyzed, which shows that the fine-tuning of the deep learning model parameters result in the overall test accuracy of 89.23% and the average classification accuracy of 91.69%.