On the time series support vector machine using dynamic time warping kernel for brain activity classification
- PDF / 1,304,262 Bytes
- 14 Pages / 595 x 842 pts (A4) Page_size
- 21 Downloads / 212 Views
ON THE TIME SERIES SUPPORT VECTOR MACHINE USING DYNAMIC TIME WARPING KERNEL FOR BRAIN ACTIVITY CLASSIFICATION1 W. A. Chaovalitwongsea and P. M. Pardalosb
UDC 612.821:51+519.7+519.8
A new data mining technique used to classify normal and pre-seizure electroencephalograms is proposed. The technique is based on a dynamic time warping kernel combined with support vector machines (SVMs). The experimental results show that the technique is superior to the standard SVM and improves the brain activity classification. Keywords: time series, classification, EEG, brain dynamics, optimization, dynamic time warping, epilepsy, support vector machines. 1. INTRODUCTION The human brain is among the most complex systems known to man. In neuroscience research, a countless number of studies attempted to comprehend the mechanism of brain functions through detailed analysis of neuronal excitability and synaptic transmission. Many theories of brain functions were proposed over the last century. Only in the last few years, it became feasible to capture simultaneous responses from sufficiently large numbers of neurons to empirically test those long-standing hypotheses about brain function. However, most neuro-scientific experiments resulted in massive datasets in the form of multi-dimensional time series data. These data contain both spatial and temporal properties of brain functions. Making sense of such massive data requires very efficient and sophisticated techniques that are capable of capturing both spatial and temporal properties simultaneously. Current research studies in data mining and classification are mostly focused on data with only spatial or temporal properties. In addition, very few studies in quantitative neuroscience are not tailored to exploit both spatial and temporal properties of this relentless flood of information. In this study, epilepsy will be a case in point. Epilepsy is the second most common brain disorder after stroke, yet the most devastating one. The most disabling aspect of epilepsy is the uncertainty of recurrent seizures that can be characterized by a chronic medical condition produced by temporary changes in the electrical function of the brain. Most epilepsy studies employ electroencephalograms (EEGs) as a tool for capturing electrical changes and evaluating physiological states (normal and abnormal) of the brain. Although EEGs offer excellent spatial and temporal resolution of brain activity, EEG data are so enormous and are represented in the form of such long-term multi-dimensional time series that neuroscientists understand very little about the dynamical transitions to neurological dysfunctions of seizures. A necessary first step to advance epilepsy research is to develop a seizure prediction/warning system. Therefore, the main goal of this study is to employ techniques in data mining and optimization to discover seizure-precursor patterns encrypted in enormous EEGs. In order to validate the reliability of a seizure prediction/warning system, one has to test the hypothesis that the EEGs during t
Data Loading...
