Distinguishing anger in the brain using machine learning on EEG data

Popescu, Mihai (2019) Distinguishing anger in the brain using machine learning on EEG data. Bachelor's Thesis, Artificial Intelligence.

Preview	Text AI_BA_2019_MihaiPopescu.pdf Download (6MB) | Preview
	Text toestemming.pdf Restricted to Registered users only Download (144kB)

Abstract

Monastic debate is a form of meditation aimed at enhancing emotional regulation. It is an integral part of monastic training and serves as a complementary practice that promotes beneficial emotions and minimizes destructive ones in the process. This study attempted at attesting the validity of this form of meditation by examining, through the use of EEG, whether monks that are experienced in monastic debate have a lower tendency of becoming angry than inexperienced monks. The expectation is that compared to beginner monks, experienced monks would have benefited more from this type of training, enhancing their ability to regulate emotions and thus exhibit a lower number of occurrences of anger. Comparing moments of anger and non-anger, we found a significant difference in terms of oscillatory power in alpha, beta, and theta frequency bands across multiple electrodes that could potentially distinguish anger. In order to be able to differentiate between anger and non anger moments on a single-trial level, three support-vector machines with different kernels and a K-Nearest-Neighbour classifier with grid search optimization were built to predict occurrences of anger using two sets of feature vectors: i) discovered by our statistical analysis ii) retrieved from literature. The best accuracies obtained were 54.52% (SVM) and 80.61% (KNN). Experienced monks were found to have shorter and less frequent anger episodes (three anger moments that lasted four seconds on average), compared to inexperienced monks (13 anger moments that lasted 7.3 seconds on average), suggesting the fact that this form of training may be able to help regulate anger. The poor accuracy scores can be improved by collecting and making use of more data, and applying different validation techniques to reduce noise.

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:	19 Aug 2019
Last Modified:	20 Aug 2019 08:49
URI:	https://fse.studenttheses.ub.rug.nl/id/eprint/20704

Actions (login required)

View Item