Inter-subject Event Related Potential Detection on Problematic Online Game Player EEG Signal Using Machine Learning Approach

• Published in: International Seminar on Research of Information Technology and Intelligent Systems (Online) pp. 784 - 788
• Main Authors: Khairani, Annisa Aulia, Ayuningtyas Kusumawardani, Sekar, Syawala, Anugrah, Wijayanto, Inung, Magdalena, Rita, Diantina, Fanni Putri
• Format: Conference Proceeding
• Language: English
• Published: IEEE 11.12.2024
• Subjects: Bayes methods; Brain modeling; Classification; EEG; Electroencephalography; ERP; Feature extraction; Games; Online Game; P300; Random forests; Scalability; Scalp; Seminars; Signal analysis
• ISSN: 2832-1456
• DOI: 10.1109/ISRITI64779.2024.10963620

Online game disorder is a condition where individuals cannot control their gaming habits in online games, leading to disruptions in their daily lives. This study utilizes inter-subject respondents among individuals with online game disorder to explore the effects of this condition. Electroencephalography (EEG) is a technique used to record the brain's electrical activity through electrodes placed on the scalp, helping to measure brain activity and analyze signals under various conditions. Stimuli are applied during recording to elicit brain responses, and Event-Related Potentials (ERP) is a method used to study cognitive processes from EEG signals. The ERP P300 appears as a positive wave with an amplitude of 6.5-20 microvolts within 200-500 ms after the stimulus is presented, and it is used to measure the brain's speed in recognizing stimuli. Classification in this study groups the data into two categories using features such as maximum peak, energy, and power, and five machine learning models are evaluated for classification. The Random Forest model achieved the highest performance with 98% accuracy, demonstrating its superiority over other models, while the Naïve Bayes model performed poorly due to dataset and feature extraction mismatches. Statistical validations, including p-values and t-tests, supported the reliability of the results. This study demonstrates the potential of machine learning, particularly Random Forest, in EEG signal analysis and emphasizes the need for future research to expand datasets and incorporate advanced models for broader applicability.