41 Comparison of Logistics Regression (LR) and Evolutionary Programming (EP) Derived Support Vector Machines (SVM) and Chi Squared Derived Results for Breast Cancer Diagnosis

Logistics Regression is used by a large community in the computer aided diagnosis (CAD) of cancer. However, very little work is presented regarding the comparison of LR to other statistical learning theory paradigms, such as SVMs. Preliminary conclusions demonstrate that both LR and EP derived SVMs do provide comparable results even though the SVM diagnostic Az is slightly more accurate. However, a surprising preliminary result is that the important mammogram discriminators of breast composition, age and family history are not significant contributors to an accurate diagnosis, as demonstrated by the Chi Squared values and verified by both diagnostic paradigms. Investigation is currently ongoing to ascertain the reason for this unexpected result.