This paper introduces an innovative approach to the approximation of large-scale dynamic systems. The primary objective is to guarantee the preservation of the original system characteristics in the reduced model. The process of deriving the reduced model involves the application of both classical and evolutionary methods. A thorough comparison is conducted, contrasting these traditional approaches with the proposed technique, which is grounded in various performance indices including the accuracy of the approximated models and the computational efficiency of the techniques. To illustrate the efficacy of the proposed methodology, three standard examples from the literature are presented. These examples vividly showcase the ease of implementation, precision, and validation of the proposed technique. The proposed methodology concludes to be a satisfactory development compared to the earlier existing techniques.