Discovering Risk of Disease with a Learning Classifier System

A learning classifier system, EpiCS, was used to derive a continuous measure of disease risk in a series of 250 individuals. Using the area under the receiver-operating characteristic curve, this measure was compared with the risk estimate derived for the same individuals by logistic regression. Over 20 training-testing trials, risk estimates derived by EpiCS were consistently more accurate (mean area=0.97, SD=0.01) than that derived by logistic regression (mean area=0.89, SD=0.02). The areas for the trials with minimum and maximum classification performance on testing were significantly greater (p=0.019 and p<0.001, respectively) than the area for the logistic regression curve. This investigation demonstrated the ability of a learning classifier system to produce output that is clinically meaningful in diagnostic classification.