Using Machine Learning to Predict Decisions on EEG Data

Kaur, Sukhleen (2019) Using Machine Learning to Predict Decisions on EEG Data. Bachelor's Thesis, Artificial Intelligence.

Preview	Text Kaur_Bachelor_Project_Thesis.pdf Download (1MB) | Preview
	Text Toestemming.pdf Restricted to Registered users only Download (122kB)

Abstract

Many studies have been conducted to determine how decisions are made in the brain. The Drift Diffusion Model (DDM) proposes an explanation for this. Studies have then tried to apply machine learning classification to determine whether decision-making processes exhibit activity as described by the DDM. Although these studies find classification accuracies above chance, they still tend to be low. Hence, this study investigates whether these results can be improved using different machine learning classification techniques. This study classifies intracranial EEG (iEEG) data collected while participants performed perceptual and memory decision making tasks. Fisher’s Linear Discriminant Analysis (LDA), a Support Vector Machine (SVM) with Radial Basis Function (RBF) kernel and an SVM with Polynomial kernel were fit onto this data and their classification was evaluated. It was found that classification was done above chance only for a few participants. The classifiers did not perform well, their F1 scores only ranged between 0.52 to 0.54 across the tasks. Polynomial SVM classified perceptual decisions best and RBF SVM classified memory decisions best compared to the other classifiers based on their ROC curves and F1-scores. The reason for a low classification rate is because of low amounts of data and because of the inconsistency of the data (difference in electrode placements). In conclusion, this study was not able to classify decisions with significant accuracy.

Item Type:	Thesis (Bachelor's Thesis)
Supervisor name:	Vugt, M.K. van
Degree programme:	Artificial Intelligence
Thesis type:	Bachelor's Thesis
Language:	English
Date Deposited:	18 Jul 2019
Last Modified:	19 Jul 2019 09:08
URI:	https://fse.studenttheses.ub.rug.nl/id/eprint/20328

Actions (login required)

View Item