The classification of Parkinson speech with feature selection by using the forest optimization algorithm

Name: Kanin Ouchin

Organization : Khon Kaen University. Faculty of Science

Email : kaniou@kku.ac.th

Name: Sirapat Chiewchanwattana

Organization : Khon Kaen University. Faculty of Science

Email : sunkra@kku.ac.th

Name: Thanaphon Tangchoopong

Organization : Khon Kaen University. Faculty of Science

Email : thanaphon@kku.ac.th

keyword: Forest Optimization Algorithm (FOA)

ThaSH: Computer algorithms

ThaSH: Machine learning -- Models

ThaSH: Parkinson's disease -- Classification

ThaSH: Computer sound processing

Abstract: Parkinson's disease (PD) is a degenerative disorder of the central nervous system. Patients with Parkinson's disease suffer from tremor, slowed movement, and speech impairment, especially dysphonia which is the most important precursor of changing in articulation and speech. So using speech data, researchers intend to classify the Parkinson's patients and healthy persons by their speech signals without invasive diagnostic. This study aims to improve the classification efficiency and reduce the attribute space of collected data. By modifying continuous search space in Forest Optimization Algorithm (FOA) for discrete search space, the proposed Feature Selection Forest Optimization Algorithm (FSFOA) iteratively selects a subset feature of the Parkinson speech dataset using the feedback from classification accuracy of machine learning models. The Machine Learning algorithms including K-Nearest Neighbors (KNN), Support Vector Machine (SVM), Decision Tree (DT) and Naïve Bayes (NB) were used to classify healthy controls from PD patients. Each classification algorithm was validated using the average classification accuracy from leave-one-subject-out (LOSO), 2-fold, and 10-fold cross-validation methods. The result shows that the best of cross-validation is LOSO; both of the classification accuracy and attribute space reduction. The KNN classifier with k parameter of 3 is the best classification accuracy of 95.24 percent, and the SVM classifier of 53.85 percent is the highest dimensionality reduction. This proposed method can be applied to other datasets.

King Mongkut's University of Technology North Bangkok. Central Library

Address: BANGKOK

Email: library@kmutnb.ac.th

Created: 2019

Modified: 2024-11-14

Issued: 2024-11-14

บทความ/Article

application/pdf

BibliograpyCitation : ใน มหาวิทยาลัยทักษิณ วิทยาเขตสงขลา. การประชุมวิชาการระดับชาติและระดับนานาชาติด้านเทคโนโลยีคอมพิวเตอร์ประยุกต์และระบบสารสนเทศ และการประชุมวิชาการระดับชาติด้านบริหารธุรกิจ ครั้งที่ 16 (ACTIS&NCOBA 16th) (pp.256-261). สงขลา : มหาวิทยาลัยทักษิณ วิทยาเขตสงขลา

eng

©copyrights King Mongkut's University of Technology North Bangkok

ลำดับที่.	ชื่อแฟ้มข้อมูล	ขนาดแฟ้มข้อมูล	จำนวนเข้าถึง	วัน-เวลาเข้าถึงล่าสุด
1	ACTIS&NCOBA 16thpp.256-261.pdf	548.88 KB

ใช้เวลา

0.02081 วินาที