Abstract
Development of robust diagnostic tests depends on improving classification accuracy of patients. Finding the thresholds for classifying the normal and the diseased is considered as an optimisation problem. For this purpose, Taguchi's signal-to-noise (S/N) ratios are applied in order to improve simultaneously the misclassification rates. Computational results of 140 diagnostic tests as well as that of a case study in female breast cancer confirmed the effectiveness of this method. Having reduced the variability in classifying the patients via Taguchi methods, more accurate and cost-efficient diagnostic tests can be designed on target in healthcare.