A Comparison between Combiner and Stacked Generalization

Combiner and Stacked Generalization are two very similar meta-learning methods that combine predictions of multiple classifiers to improve accuracy of any single classifier. Both methods form a meta-level classifier from meta-data that are predictions of multiple classifiers on the same data items. The difference between these two approaches lies in the way meta-data is formed. In this paper, we compare stacked generalization and combiner in both acurracy and efficiency. We show that both methods improve the accuracy of any single classifier roughly at an equivalent level of accuracy. But combiner's accuracy is a little higher than that of stacked generalization. This is different from general anticipation. Moreover , we also see that the cost of stacked generalization is very large and may prevent it from being used on very large data sets