It is a matter of great concern to simplify the large-scale physical systems for obtaining a better understanding of the behaviour more accurately at a faster rate. The proposed method focuses on the designing of a method of model order reduction of real time physical systems based on the mixed approach. Improved pole clustering is preferred to reduce the denominator and genetic algorithm to reduce the numerator equation. These techniques are implemented in Matlab simulation environment. The proposed order reduction technique is compared with previously designed methods. The performance comparison is done based on the calculated parameters viz. ISE, rise time, percentage overshoot, steady-state error, settling time as well gain margin and phase margin. The research work reveals that the proposed method provides an improved approximation of a large order system as compared to previous techniques with improved accuracy and better transient and steady-state response.