Adverse drug reaction (ADR) signal detection serves as a core component of pharmacovigilance, with its methodological framework continuously enriched by advancements in data science and artificial intelligence. This article systematically reviews and elaborates on the principles, implementation pathways, and application scenarios of mainstream ADR signal detection methods, including traditional frequentist methods such as the proportional reporting ratio (PRR), reporting odds ratio (ROR), and the comprehensive standard method; Bayesian probabilistic models such as the Bayesian confidence propagation neural network (BCPNN) and the multi-item gamma-Poisson Shrinker (MGPS); as well as machine learning and data mining techniques such as association rules, random forest, and zero-inflated models. By systematically comparing the advantages, limitations, applicable conditions, and empirical studies of these methods in the monitoring of both traditional Chinese medicine and chemical drugs, this study aims to provide researchers and regulatory agencies with a comprehensive methodological reference framework to facilitate the selection, optimization, and integrated application of detection approaches. Moving forward, improving data quality and promoting multi-technology collaboration will be crucial directions for enhancing the sensitivity and specificity of ADR signal detection, thereby supporting more precise decision-making in clinical medication safety.