chapter  22
Cluster Ensembles: Theory and Applications
ByJoydeep Ghosh, Ayan Acharya
Pages 20

The design of multiple classifier systems to solve difficult classification problems, using techniques such as bagging, boosting, and output combining [54, 62, 38, 36], has resulted in some of the most notable advances in classifier design over the past two decades. A popular approach is to train multiple “base” classifiers, whose outputs are combined to form a classifier ensemble. A survey of such ensemble techniques—including applications of them to many difficult real-world problems such as remote sensing, person recognition, one vs. all recognition, and medicine — can be found in [51]. Concurrently, analytical frameworks have been developed that quantify the improvements 552in classification results due to combining multiple models [61]. The extensive literature on the subject has shown that from independent, diversified classifiers, the ensemble created is usually more accurate as well as more reliable than its individual components, i.e., the base classifiers.