This paper introduced a novel order reduction method for large- scale linear dynamic single-input and single-output (SISO) systems to the lower order models. The coefficients of the reduced numerator are determined by using the Improved Pade approximation approach, while the reduced order denominator polynomial has been computed by recursive pole clustering technique. The proposed method assures the stability of the reduced order model, if the original model of high order is stable. The efficacy of the proposed method tested on few systems taken from the literature. The graphical and analytical comparison has been done with the help of MATLAB/Simulink.