AbstractSafety profiles of the influenza vaccine and its subtypes are still limited. We aimed to address this knowledge gap using multiple data mining methods and calculated performance measurements to evaluate the precision of different detection methods. We conducted a post-marketing surveillance study between 2005 and 2019 using the Korea Adverse Event Reporting System database. Three data mining methods were applied: (a) proportional reporting ratio, (b) information component, and (c) tree-based scan statistics. We evaluated the performance of each method in comparison with the known adverse events (AEs) described in the labeling information. Compared to other vaccines, we identified 36 safety signals for the influenza vaccine, and 7 safety signals were unlabeled. In subtype-stratified analyses, application site disorders were reported more frequently with quadrivalent and cell-based vaccines, while a wide range of AEs were noted for trivalent and egg-based vaccines. Tree-based scan statistics showed well-balanced performance. Among the detected signals of influenza vaccines, narcolepsy requires special attention. A wider range of AEs were detected as signals for trivalent and egg-based vaccines. Although tree-based scan statistics showed balanced performance, complementary use of other techniques would be beneficial when large noise due to false positives is expected.