Constrained Classification of Large Imbalanced （约束分类大不平衡）.pdf

International Journal of Machine Learning and Computing, Vol. 3, No. 2, April 2013 Constrained Classification of Large Imbalanced Data by Logistic Regression and Genetic Algorithm Martin Hlosta, Rostislav Stríž, Jan Kupčík, Jaroslav Zendulka, and Tomáš Hruška A. Imbalanced Data Classification Abstract—Imbalance in data classification is a frequently discussed problem that is not well handled by classical When we talk about imbalanced data, we consider data classification techniques. The problem we tackled was to learn with a significant disproportion in the occurrence of instances binary classification model from large data with accuracy of each class. Types of data imbalance are widely described constraint for the minority class. We propose a new in [1]. Experiments performed on imbalanced data sets have meta-learning method that creates initial models using shown that such type of data may often have negative impact cost-sensitive learning by logistic regression and uses these models as initial chromosomes for genetic algorithm. The on the learning phase of most standard classification method has been successfully tested on a large real-world data algorithms. The algorithm is overwhelmed by the number of set from our internet security research. Experiments prove that majority class instances which may lead to not discovering our method always leads to better results than usage of logistic the minority class. regression or genetic algorithm alone. Moreover, t