6 Performance Validation Using Several Statistical Learning Theory Paradigms for Mammogram Screen Film and Clinical Data Features Available to Purchase

In an ANNIE 2006 paper “Comparison of Logistics Regression (LR) and Evolutionary Programming derived Support Vector Machines (EP-SVMs) and CHI Squared derived results for breast cancer diagnosis”, pp267–272 of ANNIE volume 16, it was demonstrated for the first time that LR and Evolutionary Programming derived SVMs provided essentially the same diagnostic accuracy using and ROC analysis. However, in a surprising preliminary finding, it was also shown that breast composition, age and family history are not significant diagnostic discriminators. This paper extends that research by investigating the performance results, employing the same data set and statistical cross validation methods, using, in addition to the above two paradigms, the probabilistic neural network as well as a Kernelized- partial least squares approach.