作者: M Dundar , G Fung , L Bogoni , M Macari , A Megibow
DOI: 10.1016/J.ICS.2003.10.002
关键词: Test set 、 Cad system 、 Data mining 、 Classifier (UML) 、 Training set 、 Computer science
摘要: Abstract In this study we first discuss potential pitfalls involved in training a classifier for CAD system and then propose methodology successful validation of system. Our approach tries to achieve balance between performing well on the data while generalizing new cases. We performed several experiments justify each step proposed methodology. As our experimental results suggest, one can safely consider leave-one-patient-out tuning selecting relevant features as performance measure. However, final should always be evaluated an independent test set.