Understanding learning from EEG data - Combining machine learning and feature engineering based on hidden Markov models and mixed models

Aug 5, 2026

Theta oscillations in the 4–8 Hz range are closely linked to spatial learning and memory during navigation, yet scalp EEG datasets are highly complex, making behavior-related neural changes difficult to interpret. This research proposes using hidden Markov models and linear mixed effects models to engineer features from frontal theta EEG recorded during a virtual spatial navigation task. Engineered theta features from early and late trials are used to classify learner versus non-learner participants across six machine learning methods, and results are compared under different EEG standardisation approaches.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages25

Reading Time1 h 3 min

AccessFree to download

Abstract

Introduction

Spatial navigation, memory, and theta oscillations

Challenges of human scalp EEG analysis

Proposed hidden Markov and mixed-model feature extraction

arXiv :2311 .08113v1 [ q-bio .QM] 14 Nov 2023 Understanding learning from EEG data: Combining machine learning and feature engineering based on hidden Markov models and mixed models Gabriel R. Palma1,2,∗ Conor Thornberry3 Sen Commins3 Rafael A. Moral1,2 1. Hamilton Institute, Maynooth University, Maynooth, Ireland; 2. Department of Mathematics and Statistics, Maynooth University, Maynooth, Ireland; 3. Department of Psychology, Maynooth University, Maynooth, Ireland; ∗ Corresponding author; e-mail: gabriel.palma.2022@mumail.ie Keywords: Hidden Markov models, deep learning, machine learning, EEG data, time series. Manuscript type: Research paper. Prepared using the suggested LATEX template for Am. Nat. Abstract Theta oscillations, ranging from 4-8 Hz, play a significant role in spatial learning and memory functions during navigation tasks. Frontal theta oscillations are thought to play an important role in spatial navigation and memory. Electroencephalography (EEG) datasets are very complex, making any changes in the neural signal related to behaviour difficult to interpret. However, multiple analytical methods are available to examine complex data structure, especially machine learning based techniques. These methods have shown high classification performance and the combination with feature engineering enhances the capability of these methods. This paper proposes using hidden Markov and linear mixed effects models to extract features from EEG data. Based on the engineered features obtained from frontal theta EEG data during a spatial navigation task in two key trials (first, last) and between two conditions (learner and nonlearner), we analysed the performance of six machine learning methods (Polynomial Support Vector Machines, Non-linear Support Vector Machines, Random Forests, K-Nearest Neighbours, Ridge, and Deep Neural Networks) on classifying learner and non-learner participants. We also analysed how different standardisation methods used to pre-process the EEG data contribute to classification performance. We compared the classification performance of each trial with data gathered from the same subjects, including solely coordinate-based features, such as idle time and average speed. We found that more machine learning methods perform better classification using coordinate-based data. However, only deep neural networks achieved an area under the ROC curve higher than 80% using the theta EEG data alone. Our findings suggest that standardising the theta EEG data and using deep neural networks enhances the classification of learner and non-learner subjects in a spatial learning task. 1 Introduction Navigating from one place to the next is a complex cognitive skill that relies on the brain’s ability to represent spatial information and retrieve it from memory. Studies in rodents and other animals have been instrumental in uncovering foundational mechanisms of spatial cognition and memory. The hippocampus, entorhinal cortex, and parietal cortex form the core of a widespread navigation circuit. Theta oscillations in the 4-8 Hz frequency range have been shown to playa critical role in spatial learning and memory during navigation tasks. Accumulating evidence has demonstrated the role of frontal midline theta in spatial learning and exploration (Chrastilet al., 2022a; Crespo-Garca et al., 2016; Du et al., 2023a; Liang et al., 2021a; Roberts et al., 2013; Thornberry et al., 2023) as well as successful retrieval (Buzski, 2005; Greenberg et al., 2015; Herweg et al., 2020; Kaplan et al., 2014, 2012; Klimesch et al., 1997; Lin et al., 2017; Robertset al., 2013) . It is possible that frontal theta oscillations facilitate communication between the hippocampus and the cortex to support the encoding of spatial memories (Buzski, 2005; Buzski and Moser, 2013; Herweg et al., 2020; Kerrn et al., 2018; Liang et al., 2021a; Mitchell et al., 2008) . However, analysing human scalp-EEG data collected during real-worl

FAQs

Why are frontal theta oscillations important for spatial learning tasks?

Frontal theta oscillations within the 4–8 Hz range are associated with spatial learning and memory during navigation, supporting processes such as exploration and successful retrieval.

How does the paper extract features from EEG data?

It uses hidden Markov models together with linear mixed effects models to engineer informative features from frontal theta EEG data from key trials under different conditions.

Which model and input combination achieved the best classification performance?

Using theta EEG data alone, only deep neural networks produced an ROC AUC above 80%, and standardising theta EEG data improved learner vs non-learner classification.

Aug 5, 2026

Upload by Ava Thompson