Construction and Evaluation of a Robust Classification Model for Multi-objective Pr oblems
Hung‐Yi Lin, Yu-Han Lai · 2011
Abstract—Classification by using of multiple variables is a frequently encountered data mining problem. Conventional classification models can either suffer from insufficient data collection or be burdened with overabundant data. In this paper, we propose a novel model in generating a robust multivariate classifier to solve the overabundance case. The classification problems with multiple objectives can be supported by a subset of effective variables identified by our scheme. Traditional Gini index and principal component analysis (PCA) are integrated to complete our classification model. Some experiments based on practical databases are conducted to verify the robustness of our method. Index Terms—Multi-objective classification, Gini index, PCA, multivariate classifier