Date of Award

2007

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Bioinformatics & Computational Biology

First Advisor

Dr. Remzi Seker

Abstract

In this study, we try to determine validity of the several different measures applied to the nonlinear HRV data. There were many studies done using traditional methods of Approximate Entropy, Correlation Dimension, Lyapunov exponents and recurrence plots. While employing some of these methods, we also decided to use Support Vector Machines. This method of analysis is quite new to the world of nonlinear data. We obtained data from five different groups of young healthy adults: spontaneous breathing group (Normal), metronomic breathing group (Metron) and elite athletes (Ironman), as well as two groups of people participating in specific traditional forms of Chinese Chi and Kundalini Yoga meditation techniques. We have applied ApEn, CD, CTM and SVM methods to the data sets of equal length, around 4000 points each. Mean and standard deviation were used on data to help analysis using traditional methods. We applied methods of Recurrence Quantification Analysis to recurrence plots to analyze the images we received. Finally, we used six different kernel functions and applied Lagrange multipliers, margin and percentage of vectors used to identify which of the SVM kernel function are useful in the classifying of HRV data. As a result of our study, we discovered that not all of the methods listed about are useful in classifying such a data. Traditional types of analysis most of the time struggled with data that was not linearly separable. However, once the degree of separability increased, we were able to classify it using both traditional methods, as well as SVM kernel functions.

Download

Share

COinS