The Effect of Visual Stimulus Type on Machine Learning Model Performance in Alcoholism Detection Using EEG Signals

Ethanol is widely used in many communities, and long-term exposure can lead to dependence and alcohol-related disorders. These conditions are often linked to changes in brain activity. Electroencephalography (EEG) records the electrical signals generated by neural activity and provides a way to record these changes by converting these signals into a form that can be understood. Because EEG is accessible and practical, it has become a popular method for studying how alcohol affects the brain. A number of studies have compared EEG recordings from individuals with alcohol use disorder and healthy participants. These works have used frequency-based features, entropy measures, and statistical descriptors to describe differences between the two groups. Their findings show that alcohol use can influence the brain's response to sensory input and cognitive tasks. Researchers haven't really looked closely at whether the type of visual stimulus matters when trying to tell two groups apart. There's a basic question here that still needs answering: can showing just one picture do the job, or does flipping through two images back-to-back give us a clearer picture of what makes these groups different? Getting to the bottom of this could actually help scientists design better EEG experiments and make more sense out of the data they collect. The dataset used in this study was collected by Zhang and colleagues and is available in the UCI Machine Learning Repository. EEG signals were recorded from sixteen participants, including eight individuals with alcohol use disorder and eight healthy controls. Recordings were made under three conditions: a single image shown to the participant (S۱ obj), two identical images presented consecutively (S۲ match), and two different images shown in sequence (S۲ nonmatch). The aim of this study is to compare the classification performance across these three stimulus conditions. A broad set of features commonly used in EEG research was extracted to evaluate the differences between the groups. By examining how each condition influences classification accuracy, this work seeks to identify which stimulus type provides clearer separation. The results can support researchers in selecting more effective stimulus designs when planning EEG studies related to alcohol use.

کلیدواژه ها:

Electroencephalography ، Alcoholism ، Visual Stimulus ، Feature Extraction ، XGBoost ، Classification