Analysing the Performance of LDA (Linear Discriminant Analysis) Feature for Diagnosing PAF (Paroxysmal Atrial Fibrillation) Patients


Sadaghiyanfam S., KUNTALP M.

International Scientific Meeting on Electrical-Electronics and Biomedical Engineering and Computer Science (EBBT), İstanbul, Turkey, 24 - 26 April 2019 identifier identifier

  • Publication Type: Conference Paper / Full Text
  • Volume:
  • Doi Number: 10.1109/ebbt.2019.8741871
  • City: İstanbul
  • Country: Turkey
  • Keywords: linear discriminat analysis, HRV, paroxysmal atrial fibrillation, normal sinus rythm, ECG, support vector machine, SUPPORT VECTOR MACHINES, ONSET
  • Dokuz Eylül University Affiliated: Yes

Abstract

Linear Discriminant Analysis (LDA) is an offered scheme for extracting feature and reducing dimension. LDA has been extensively utilized in various applications including high-dimensional dataset. In this study, we analyzed the effectivity of LDA feature extracted from 33 short-term Heart Rate Varibility (HRV) ECG records for diagnosing of Paroxysmal Atrial Fibrillation (PAF) from normal sinus rhthm (NSR) ECG records. By this way, one set of feature obtained from LDA was utilized as a new input to the classification which is chosen as the Support Vector Machine (SVM). The results represent that dimesnion reduction by LDA effects negatively on accuracy, seneistivity, precision and specificity. In contract, increase the perfromance in term of AUC.