Developing a scorecard using a simple artificial immune systems (SAIS) algorithm and a real-world unbalanced dataset

A simple artificial immune system (SAIS), which was previously developed, can predict class outcomes accurately and therefore has good classification accuracy, which is the percentage of correctly classified data. Classification accuracy works well on balanced datasets; however, since in this study, a large unbalanced dataset was obtained, classification accuracy cannot be used as a measure of performance. Instead, the Gini coefficient, which is the main performance measure used in industry for generating scorecard and which is insensitive to changes in class distribution, will be used. SAIS was modified to generate a Gini coefficient and an investigation of its suitability for scorecard development was made. We found that further modifications are needed in order for it to perform as well as logistic regression, which is the main technique used in practice for developing scorecard.