Using Falsity Data in the Stacking Technique
Pawalai Kraipeerapun, Somkid Amornsamankul · 2016
This paper proposes the use of original falsity datain the stacking technique. Instead of using the average betweenthe truth and non-falsity data predicted from level 0, theoriginal truth data and the original falsity data predicted fromlevel 0 are applied to level 1 of the stacking. In the experiments,the proposed technique is used to solve both classification andregression problems from UCI machine learning repository. Itis found that the performance of the proposed technique isbetter than the traditional stacking technique.