We compared the accuracy of 3 data-mining models, neural-network, decision-tree, and logistic-regression, in predicting the 5-year survival of patients with colorectal cancer. The database consisted of patient demographics, pathologic features, and levels of expression of 2 biomarkers (p53 and Bcl-2). All 3 methods demonstrated acceptable accuracy, from 64% to 70%. The neural-network model had the best specificity (80%) and accuracy (70%) but lowest sensitivity (59%). Both logistic-regression and decision-models demonstrated comparable sensitivity (72%).