Prediction of Neurodegenerative Diseases Based on Gait Signals Using Supervised Machine Learning Techniques

摘要

In recent year’s neurodegenerative diseases such as Amyotrophic Lateral Sclerosis (ALS), Huntington’s disease (HD), and Parkinson’s disease (PD) has gain lot of attention because the money spend on the healthcare services generated by these diseases has created a hugeburden in the economy of the developed countries. These diseases are progressive in nature and it produces some dead cells in the brain and spinal cords that directly affect the gait, speech and memory loss. As the old age population is increasing at a higher rate, it is necessary to developsome intelligent technique to detect these diseases at the early stage to reduce the economic burden. With the advent of machine learning techniques, classification based on gait signals has become popular these days. Few past research have been made to classify the neurodegenerative diseasesby using binary classification approach with linear feature selection algorithm and linear classifiers. In this paper quad classification approach was used by considering all groups (ALS, HD, PD and Control) with Recursive Feature Elimination (RFE) algorithm for feature selection and usingsupervised machine learning techniques such as linear, nonlinear classifiers with decision tree and probabilistic classifiers. Finally the performance measures of each classifiers has been studied with 5 selected features obtained from RFE method and was found that the nonlinear classifiersuch as Random Forest and Bagging CART have given best performance with an accuracy of 96.93% and 97.43% respectively. This analysis helps the clinicians to distinguish neurodegenerative disease from the healthy group by using gait signals.