<p>In the context of large-scale datasets and complex tasks, traditional classification methods often face challenges such as data imbalance, noise interference, and feature selection. Ensemble classification methods, as an important approach to address these issues, combine the strengths of multiple classifiers and have achieved significant results in many practical applications. However, how to resolve these challenges according to the different characteristics and requirements of the data remains a key challenge in ensemble classification research. To introduce the current state and challenges of ensemble classification research, this paper reviews the literature on ensemble classification from the dual perspective of sampling strategies and learning methods. For the first time, this paper provides a detailed analysis and summary of ensemble methods based on different sampling strategies, including data optimization sampling strategies such as uncertainty sampling, instance selection, and stratified sampling, as well as data adjustment sampling strategies like hybrid sampling, oversampling, and undersampling. The performance of various algorithms based on different datasets is compared using various evaluation metrics. Additionally, learning strategies based on auxiliary information such as meta-learning, transfer learning, and semi-supervised learning, as well as dynamic optimization learning strategies based on incremental learning, online learning, and cost-sensitive learning, are systematically analyzed and summarized. The time complexity of classic algorithms is also examined. Finally, the advantages and limitations of each strategy are summarized, and future research directions in ensemble classification are proposed, with an emphasis on selecting appropriate strategies to improve classification performance in practical applications.</p>