抖音作为中国领先的短视频平台，通过提供沉浸式和高度互动的内容体验，重塑了固有的社交互动模式，但有关其认知层面的影响的研究仍然有限。本研究填补了针对短视频平台播放量表现不同的短视频之间的脑电波差异的研究空白，并探索了使用脑电（EEG）数据预测视频发布前表现的可能性。本研究使用自制的模拟短视频应用的播放器，记录了30名参与者在观看短视频内容时的脑电数据。研究分析了自我报告的偏好评分、观察到的大脑活动与视频表现数据之间的相关性。另外，采用了包括最近邻算法（kNN）、随机森林和XGBoost在内的机器学习算法，探索了预测视频表现的可行性。研究结果显示，播放量表现好的视频通常会引起较低的Alpha波、Theta波活动，且较高的Beta波活动，说明观众的参与度和注意力水平较高。此外，我们还发现Beta波活动与视频表现之间存在显著的正相关关系，这也进一步确认Beta波活动可能是观众喜好的一个潜在指标。机器学习模型在视频的播放量预测上呈现了较高的准确率，对于播放量为100万以下的视频，准确率高达86.91%，对于播放量区间为100万到300万次的视频，准确率高达78.49%，从而表明，利用EEG数据预测尚未发布的短视频的表现具有一定的可行性。
Douyin, China's leading short-video platform, has reshaped social interaction by providing users with immersive and highly interactive content experiences, yet research on its cognitive effects remains limited. This study fills a research gap by exploring neurophysiological differences between high-performing and low-performing short videos on the Douyin platform and explores the potential of using electroencephalogram (EEG) data to predict video performance prior to its release.Utilizing a custom video player, EEG responses from 30 participants were recorded as they interacted with short video content. This study analyzed correlations between self-reported preference scores, observed brain activity, and video performance data. Machine learning techniques, including k-Nearest Neighbors (kNN), Random Forest, and XGBoost, were employed to assess the feasibility of predicting video performance.Results indicated that high-performing videos consistently elicited lower alpha and theta band activities but higher beta band activities, suggesting higher levels of viewer engagement and attention. Additionally, a significant positive correlation was found between beta activity and video performance, reinforcing beta activity as a potential indicator of viewers’ likability. The predictive models demonstrated high accuracy, reaching up to 86.91% and 78.49% accuracy for videos with traffic range with less than 1 million views and within 1 to 3 million views respectively, thereby validating the feasibility of using EEG data in predicting the likeability and performance of short videos.These findings enhance understanding of the cognitive impact of short-video content and underscore the potential of EEG data in media analytics, paving the way for future research on digital media consumption.
