A hybrid methodology for the mahalanobis-taguchi system using random binary search-based feature selection

The Mahalanobis-Taguchi system (MTS) is a relatively new statistical methodology combining various mathematical concepts and is used in the field of diagnosis and classification in multidimensional systems. MTS is a very efficient method and has already been applied to a wide range of disciplines. However, its feature selection phase (optimization stage), which uses experimental designs (orthogonal array, OA), is susceptible to improvement. In MTS, selection of important features or variables to improve classification accuracy is done using signal-to- noise (S/N) ratios and OA. OA has been noted for limitations in handling a large number of variables. Therefore, in this research, we propose the use of a random binary search (RBS) algorithm incorporated within MTS for optimizing the procedure for selecting the most useful variables. Ensemble is a powerful technique to achieve improvement in the accuracy of predictive models, whereby individual methods, which are not consistently the best performers in different problems and datasets, are brought together to provide predictions which are more accurate than those made by individual methods.