Research for multidimensional systems diagnostic analysis based on improved mahalanobis distance

The relative importance of variables is different when comprehensively evaluating samples from multidimensional systems; however, traditional Mahalanobis-Taguchi System (MTS) method ignores this difference and treats all variables equally. To improve the accuracy of diagnosis or prediction when using MTS to measure the degree of abnormality of multidimensional observations, this paper makes an improvement on Mahalanobis distance function. Firstly, the necessity and feasibility of weighted Mahalanobis distance are analyzed. Secondly, the weighted Mahalanobis distance function in MTS is proposed. Finally, based on the new method a blood viscosity diagnosis system is optimized and analyzed and the result proves the effectiveness of this method.