Analyzing EEGs Through Dimensionality Reduction for EEG Classification

Overview

“Thought-Code”/” Blink-Code” Part II: Analyzing EEGs through dimensionality reduction for EEG classification Sarvesh R. Babu duPont Manual High School Abstract The focus of the project is to identify the best dimensionality reduction methods for classification of EEG (Electroencephalographic) data. Two possible applications of EEGs are administering AEDs and Cybersecurity. The problem with AEDs is that there are many different AEDs and these drugs are not successful in all patients. No patients have the time or endurance to try every drug available, and eventually, prefer to ‘accept’ their seizures. The application in cybersecurity is extremely easy to see. EEG classification is promising for cybersecurity as it is extremely complex and hard to predict. Dimensionality reduction is a necessary step in classifying EEG data because EEG data is high dimensional, which could lead to overfitting or long computation times. The purpose of this study is to find whether Principal Component Analysis (PCA) or Linear Discriminant Analysis (LDA) is more effective at reducing the dimensions of EEG data. The methods that were used in the study were the application of PCA and LDA algorithms on EEG data sets. The codes were designed and ran in Matlab. The data analysis yielded several conclusive data sets. The reduced LDA data set and the reduced PCA data set. The reduced PCA datasets contained around 12.5 dimensions while holding 92.55% of the original variance of the data. The reduced LDA dataset contained 1 dimension while withholding 99% of the variance. The results conclude that LDA is the more efficient algorithm for reducing EEG data dimensions, but more data is needed to be tested through a classification algorithm to test true efficiency. Keywords: LDA, EEG, PCA, Dimensionality Reduction, Classification

Competition history

• AJAS 2019 Category not listed

Related projects

Closest projects by meaning, across every fair and year in the corpus.

Browse more like this

• All Computational Biology and Bioinformatics projects
• AJAS Computational Biology and Bioinformatics projects
• Every AJAS 2019 project