摘要：Ensemble Methods grows along with Machine Learning and Computational Intelligence domain proves to be effective and versatile. It helps in reducing variance and improves accuracy. Few Machine Learning challenges such as data stream classification, class Imbalance in datasets and occurrence of concept drift in non-stationary environments is addressed effectively by Ensemble methods. Data stream refers to rapidly generated heterogeneous data in a continuous way. One of the key challenges considered in learning from data streams is the detection of concept drift, i.e., changes in data distribution underlying data streams, observed over time. Such changes in incoming data deteriorate the accuracy of the classifier since classifier has been learned over past data instances that are stable. Thus detection of concept drift is an important task. The real life examples of drift are spam detection, credit card fraud detection, and weather predictions. This paper presents a survey on highlighting recent research ideas in concept drift using ensemble methods and also provides a comprehensive introduction to ensemble methods, data stream classification models, types of concept drift and drift detection methods.
