An Early Detection of Parkinson’s Disease using Machine Learning and Deep Learning Models

The human nervous system is the controlling mechanism for coordinating all body-related activities. In the cognitive world, it stimulates thinking capabilities. It prompts a reaction to a particular situation and allows body parts to move. In addition, it also regulates other physiological functions such as digestion and respiration. The nervous system consists of billions of cells which interpret information from sensory elements and translate it into coordinated action performed by various parts of the body. With the increase in age, the steady decline of these cells results in a loss of dopamine compounds. Parkinson's disease (PD) is caused by an unforeseen expansion of the movement of bodily extremities that results in this decline. Finding a cure for this disease is still an enormous challenge, and its progression can be slowed by early detection. The proposed work has been implemented with seven machine learning and deep-learning techniques, including decision tree (DT), random forest (RF), logistic regression, naive Bayes, k nearest neighbour, neural network model, and support vector machines for preemptively predicting PD. The model was implemented utilizing a dataset acquired from Kaggle. This included the normal cases as well as those with PD. The findings showed that the random forest model was superior to others, achieving an accuracy of 86.70%. In contrast, the naive Bayes model displayed the lowest accuracy of 71.50%. In addition, both the random forest and support vector machine models showed the highest accuracy levels of 85.90%. Various other performance metrics such as computational time, AUC (the Area Under the Curve), F1-score, ROC curve area, and recall rate were utilized to predict the effectiveness of each technique. Comparative analysis identified that the least performing models are the k-nearest neighbour and naive Bayes methods. In the future, this method can be refined further by using a large dataset and employing deep learning models for higher predictive outcomes.