J.B. Pugliesi, R.A. Sinoara, and S.O. Rezende (Brazil)
Data Mining, Machine Learning, Combination of Regres sors, Bagging, Boosting, Stacking.
With the objective of improving the precision of the knowl edge extracted from large databases, a combination of ho mogeneous and heterogeneous regressor can be accom plished. The bagging, boosting and stacking methods per form this combination. In this article an implementation of the bagging and boosting method is presented. To the boosting implementation a strategy based on a threshold calculated using the standard deviation of rules error, to de termine which example is correctly predicted, is used. Ex periments with three databases were run to verify the error within the knowledge obtained by combined regressors.
Important Links:
Go Back
