大脑胶质瘤诊断中不均衡问题上的特征选择

• 作 者: 刘天羽 李国正
• 机 构: 上海电机学院电气学院 上海200240 同济大学控制科学与工程系 上海201804
• 出 处: 第三届中国数据挖掘学术会议(CCDM2009)
• 发表日期: 2009年1期
• 关键词: feature selection imbalanced data sets ensemble learning mutual information dies

　　The degree of malignancy in brain glioma dominates the way of treatment and is critical before brain surgery.Many learning methods are used like rule induction algorithm, single neural networks, support vector machines, etc.But, class imbalance problem is encountered in diagnosing brain glioma, which causes seriously negative effect on the performance of classifiers that assume a balanced distribution of classes.Though it is critical, few previous works paid attention to this class imbalance problem in prediction of malignancy degree in brain glioma.There are many labeled data sets which have an unbalanced representation among the classes in them.When the imbalance is large, classification accuracy on the smaller class tends to be lower.In particular, when a class is of great interest but occurs relatively rarely such as cases of fraud, instances of disease, and so on, it is important to accurately identify it.In imbalanced problems, some features are redundant and even irrelevant.These features will hurt the generalization performance of learning machines.Here we propose MIEE (Mutual Information based feature selection for EasyEnsemble) to solve the class imbalanced problem.MIEE is used to predict the degree of malignancy in brain glioma.Experimental results on UCI data sets and asset of brain glioma data show that MIEE improves the classification performance and prediction ability on the imbalanced dataset.