Hierarchical ensemble learning method in diversified dataset analysis

Abstract The remarkable advances in ensemble machine learning methods have led to a significant analysis large data, such as random forest algorithms. However, the algorithms only use current features during process of learning, which caused initial upper accuracy’s limit no matter how well are. Moreover, low classification accuracy happened especially when one type observation’s proportion is much lower than other types training datasets. aim present study design hierarchical classifier try extract new by regressors and statistical inside whole process. In stage 1, all categorical variables will be characterized algorithm create variable through regression while numerical left serve sample factor (FA) calculate factors value each observation. Then, learned 2. Diversified datasets consist used method. experiment results show that increased 8.61%. Meanwhile, it also improves observations with dataset significantly.