In this paper, a study of the effect of a group of transformer measurements on Health Index (HI) calculation is presented. Different methodologies for selecting the most efficient group of diagnostic measurements used in classifying transformer HI are investigated. A Binary Cat Swarm Optimization (BCSO) technique is undertaken based on Support Vector Machines (SVM). The technique depends on selecting the optimal parameters for SVM. The effect of selecting HI classes as well as classs boundaries is also studied. The measurements of fourteen diagnostic transformer tests, including the furan analysis, dissolved gas analysis, and further oil analysis for 724 distribution transformers are studied, and the corresponding HI is calculated according to industrial standards. The model renders the best-selected group of measurements that assist in the formulation of the health index with minimum error and high confidence.