Stacking学习与一般集成方法的比较研究

Ensemble learning has been widely used because of the great generalization ability by combining multiple learners. The general methods such as AdaBoost, Bagging are usually based on the same algorithm, but Stacking is based on different algorithms. Aiming at classification problems, this paper constructed a framework for Stacking consist of 5 types of base classifiers which are naive bayes, logistic regression, k nearest neighbor, decision tree and rule learning. And it made a comparasion with 5 other methods that are AdaBoost, Bagging, random forests (RF), voting plan and selecting the best using cross validation. The experiments of 2 simulated and 36 real datasets were conducted and the results showed that Stacking is the best among all others according to the claasification presion and is more suitable for large sample cases.