Optimization on reference space of Mahalanobis-Taguchi System based on hybrid encoding genetic algorithms

The Mahalanobis-Taguchi System (MTS) is a new classifying and diagnostic technique of pattern recognition using a collection of methods of Mahalanobis distance, orthogonal arrays and signal-to-noise ratios. The advantages of MTS, such as the ability of choosing the effective variables, no need to make assumptions about data distribution and high classification speed, etc., make it a wide range of applications in areas, including industrial production, business management and pattern recognition. However,as a relatively new classification method, it is also controversial to construct a reference space by using the method of orthogonal arrays and signal-to-noise ratios. For the shortage of MTS, an optimization model of reference space is constructed by hybrid encoding genetic algorithms, built decision model which aimed to minimize the error rate of classification and maximize the signal-to-noise ratio, then give the initial solution to the problem by using uniform distribution strategy, and then the method of genetic algorithm is introduced to obtain further improvements. The experiment results show the good convergence and effectiveness of the model.