Enhancing Extraversion Classification With Sample Entropy: A Comparison of Two EEG Epoch Lengths

Abstract

With the advancement of technology, many researchers have begun to employ electroencephalography (EEG) to assess extraversion personality traits, replacing subjective self-report questionnaires. However, most EEG studies are time-consuming and have inadequate classification accuracy. Thus, this letter proposes a framework for extraversion classification using sample entropy features extracted from resting-state EEG signals. The proposed framework compares two different EEG epoch lengths (15 and 120 s) and evaluates their impact on classification performance. To enhance the classification performance, a sequential forward selection method is applied to ensure that only the most optimal features are utilized. Using support vector machine, k-nearest neighbors, random forest, and extreme gradient boosting as classifiers, the study shows that sample entropy outperforms power and coherence features in classifying extraversion. Remarkably, the framework achieves 100% classification accuracy using a single feature: the sample entropy from a 15-s eyes-open condition at the Fpz electrode. By reducing the number of required features to just one and focusing on a shorter EEG epoch length, this finding reflects the potential of developing EEG-based sensor systems that are more practical and cost-effective in the future.