AUTHOR=Laufer Ilan , Mizrahi Dor , Zuckerman Inon 

TITLE=Enhancing EEG-based attachment style prediction: unveiling the impact of feature domains

JOURNAL=Frontiers in Psychology

VOLUME=Volume 15 - 2024

YEAR=2024

URL=https://www.frontiersin.org/journals/psychology/articles/10.3389/fpsyg.2024.1326791

DOI=10.3389/fpsyg.2024.1326791

ISSN=1664-1078

ABSTRACT=Attachment styles, integral to human relationships, have garnered attention in the context of neurophysiological responses and EEG data analysis. We examined EEG data's potential to predict and differentiate secure and insecure attachment styles, shedding light on the neural underpinnings of interpersonal dynamics.Our study included 27 participants, applying an XGBoost classifier to diverse EEG feature domains encompassing time-domain, complexity-based, and frequency-based attributes. Results unveiled precision disparities in attachment style prediction, with an impressive 96.18% precision for insecure attachment predictions and a lower 55.34% precision for secure attachment predictions. This discrepancy underscores the intricacy of utilizing EEG patterns for attachment style predictions, influenced by the spectrum-like manifestation of attachment insecurities.Individuals with heightened perceived insecurity consistently fell within the insecure attachment group, reflecting their increased emotional reactivity and social cue sensitivity. We also emphasized specific EEG feature domains, with time-domain features playing a pivotal role in prediction accuracy, followed by complexity-based features. While informative, frequency-based features had a more modest impact. Balanced accuracy metrics revealed an overall accuracy of around 84.14%, highlighting the model's ability to classify attachment styles while considering dataset imbalances.Despite the nuanced nature of attachment styles, our research enhances our understanding of the neural correlates of attachment and suggests avenues for future exploration, including diversifying demographic cohorts and integrating data from multiple modalities to improve predictive models.