Predicting the prognosis of cardiac arrest patients is important in medical decision-making and patient treatment. Early prognosis prediction leads to appropriate medical interventions, ultimately improving patient outcomes. Therefore, developing an automated method that can reliably predict the prognosis of cardiac arrest patients at an early stage is highly valuable in the medical field. In this paper, we propose novel methods for predicting neurological recovery from coma after cardiac arrest by utilizing the dominant information flow in multi-channel EEGs. Conventional methods for analyzing neural activity mainly rely on a few EEG channels independently, and most of the multi-channel EEG-based studies are limited to specific regions of the brain. This limitation results in the inadequate reflection of the function of the cerebral cortex over wide brain regions in neurological recovery. To address this limitation, we measure neurological recovery by quantifying the dominant information flow obtained from the minimum information bipartition (MIB). We use the MIB method to detect the dominant information flow in the entire EEG channels and quantify it by calculating transfer entropy (TE). Moreover, we apply a filter-bank technique to investigate the time evolution of information flow in four frequency bands, reflecting the changes in EEG activation bands according to the patient's condition. This study utilized the subject's age and gender, as well as the transfer entropy values obtained from each frequency band as features, with the support vector machine (SVM) applied as the classifier. We performed cross-validation using 100 randomly generated classification models. Although we were able to confirm a score of 0.91 in our self-evaluation, the score we obtained in the challenge was 0.16. Our team name for the Physionet Challenge 2023 was 'ATsens'