A Survey on Concept Drift Adaptation This survey presents an overview of concept drift adaptation in online supervised learning, where the relationship between input data and the target variable changes over time. The paper discusses existing strategies for handling concept drift, evaluates adaptive learning algorithms, and presents illustrative applications. It aims to unify concepts and terminology among researchers and provide a comprehensive state-of-the-art review of methodologies and techniques for concept drift adaptation. Concept drift refers to changes in the conditional distribution of the target variable given the input features. It can be categorized into real concept drift (changes in $ p(y|X) $) and virtual drift (changes in $ p(X) $ without affecting $ p(y|X) $). The paper discusses various types of drift, including population drift, real concept drift, and virtual drift, and their implications for predictive models. Adaptive learning algorithms are designed to update predictive models online to react to concept drift. These algorithms can be classified based on their memory and forgetting mechanisms, data management strategies, and change detection techniques. The paper presents a taxonomy of methods for concept drift adaptation, including data management, change detection, learning, and loss estimation. The paper also discusses the requirements for predictive models in changing environments, such as detecting concept drift, adapting to changes, and being robust to noise. It presents an online adaptive learning procedure, where models are updated incrementally based on new data. The paper provides illustrative applications of concept drift adaptation in various domains, including monitoring and control, management and strategic planning, personal assistance and information, and ubiquitous environment applications. These applications highlight the importance of adaptive learning in handling concept drift in real-world scenarios. The survey concludes with a discussion of the taxonomy of methods for concept drift adaptation, emphasizing the importance of modular components in adaptive learning systems. The paper provides a comprehensive overview of the state-of-the-art techniques and methodologies for handling concept drift in online supervised learning.A Survey on Concept Drift Adaptation This survey presents an overview of concept drift adaptation in online supervised learning, where the relationship between input data and the target variable changes over time. The paper discusses existing strategies for handling concept drift, evaluates adaptive learning algorithms, and presents illustrative applications. It aims to unify concepts and terminology among researchers and provide a comprehensive state-of-the-art review of methodologies and techniques for concept drift adaptation. Concept drift refers to changes in the conditional distribution of the target variable given the input features. It can be categorized into real concept drift (changes in $ p(y|X) $) and virtual drift (changes in $ p(X) $ without affecting $ p(y|X) $). The paper discusses various types of drift, including population drift, real concept drift, and virtual drift, and their implications for predictive models. Adaptive learning algorithms are designed to update predictive models online to react to concept drift. These algorithms can be classified based on their memory and forgetting mechanisms, data management strategies, and change detection techniques. The paper presents a taxonomy of methods for concept drift adaptation, including data management, change detection, learning, and loss estimation. The paper also discusses the requirements for predictive models in changing environments, such as detecting concept drift, adapting to changes, and being robust to noise. It presents an online adaptive learning procedure, where models are updated incrementally based on new data. The paper provides illustrative applications of concept drift adaptation in various domains, including monitoring and control, management and strategic planning, personal assistance and information, and ubiquitous environment applications. These applications highlight the importance of adaptive learning in handling concept drift in real-world scenarios. The survey concludes with a discussion of the taxonomy of methods for concept drift adaptation, emphasizing the importance of modular components in adaptive learning systems. The paper provides a comprehensive overview of the state-of-the-art techniques and methodologies for handling concept drift in online supervised learning.