Abstract: ‍ ‍In recent years， danmaku videos have emerged as a rapidly developing form of new media communication that has gained widespread popularity. However， according to surveys， some users believe that danmaku videos negatively affect the video viewing experience， leading to suboptimal quality of experience （QoE）. As such， the evaluation of QoE has become an urgent issue in the efforts to improve danmaku video services. Existing QoE evaluation methods are limited by subjective factors that can lead to false feedback， untimely feedback， and difficulties in measuring data relationships. To address these limitations， electroencephalography （EEG） has been preliminarily applied to subjective evaluation of audiovisual stimuli due to its advantages， such as its inability to be disguised， high temporal resolution， and greater statistical significance of data. In light of these advantages， this study proposed a breakthrough use of EEG in the QoE evaluation of danmaku videos. Based on phase locking value， functional connectivity feature brain networks were constructed， and paired differential networks were extracted at two different QoE levels （high/low）. Machine learning methods were then employed to construct an evaluation model based on this network， with an average classification accuracy of 80%. The study revealed changes in the cooperation patterns of brain regions that produce different QoE in users of different types of videos and found the frontal lobe， highly related to QoE， as the main convergence area. The above research results indicate that the evaluation model can accurately record the physiological and psychological activities of users when watching videos， and provide a neurophysiological basis for improving danmaku video services.