Abstract
James Parkinson described Parkinson’s disease as a neurological syndrome, it affects the central nervous system as a result the patients face the issue in talking, strolling, tremor during motion. Parkinson’s disease patient typically has a low-volume noise with a monotone quality; this system explores the classification of audio signals feature dataset to diagnosis Parkinson’s disease (PD), the classifiers we utilized in this system are from Machine Learning. The significant classification models that are effectively utilized are (K-Nearest Neighbor) KNN, Decision Tree, Logistic regression and eXtreme Gradient Boost (XGboost). The performance analysis of the Parkinson’s disease datasets has a variety of features. There are 21 features which play an important role for predicting the best algorithm in the classification. The system has achieved better result in predicting the PD patient healthy or not, XGboost provided the peak accuracy of 92% and the Matthews Correlation Coefficient (MCC) of 78%. MDVP: Jitter (%), Jitter (Abs), MDVP: RAP, MDVP: PPQ, Jitter: DPP, MDVP: Shimmer, MDVP: Shimmer (db), Shimmer APQ3, Shimmer APQ5, MDVP: APQ, Shimmer: DDA, NHR are the selected features from the dataset, which are important for the prediction to achieved high accuracy.
Description
For analysis, the common features considered are jittering, shimmering and voice signal frequency. Various algorithms for machine learning have been used in the prediction of Parkinson's disease. The range of prediction accuracy from applied algorithms lies between 75% and 92%, which demonstrates that machine learning algorithms can be used to predict Parkinson's disease by analyzing voice signals. Here the classification accuracy was studied and compared, with good performance and fast implementation, XGboost achieved a high accuracy with 92%. This system provides the comparison between machine learning classifiers of Logistic regression, Decision tree, K-Nearest neighbor and XGboost in Parkinson’s disease diagnosis with high dimensional data MDVP: Jitter (%), Jitter (Abs), MDVP: RAP, MDVP: PPQ, Jitter: DPP, MDVP: Shimmer, MDVP: Shimmer (db), Shimmer APQ3, Shimmer APQ5, MDVP: APQ, Shimmer: DDA, NHR are the selected features from the dataset to achieve high accuracy in prediction.
Links and resources
Tags