Credit assessment involves predicting applicant reliability and profitability. The objective of this paper is to apply a number of algorithms to credit-card assessment. Despite the fact that many numerical and connectionist learning algorithms address the same problem of learning from classified examples, very little is known regarding their comparative strengths and weaknesses. Experiments comparing the top-down induction-learning algorithms (G&T and ID3) with the multilayer perceptron, pocket, and back-propagation neural learning algorithms have been performed using a set of approved applications for credit cards from the Bank of Scotland where the decision process was principally a credit scoring system. Overall, they all perform at the same level of classification accuracy, but the neural algorithms take much longer to train. The paper describes our motivation for using the machine-learning algorithms for credit-card assessment, describes the algorithms in detail, and compares the performance of these algorithms in terms of their accuracy. © 1992 Oxford University Press.