<p>The stable operation of modern accelerator-based light sources critically depends on precise optical correction to achieve high beam quality and overall machine performance. In addition, advanced machine learning algorithms require accurate linear optics parameters to construct reliable models for accelerator systems. This paper presents a method for linear optics correction based on global orbit data, which has been successfully implemented at the Shanghai Soft X-ray Free-Electron Laser. By employing linear optics analysis and transfer matrices, this approach enables the effective identification and correction of optical parameter errors arising from various sources, thereby mitigating beam orbit distortions and instabilities, and enhancing overall accelerator performance. Experimental results demonstrate that global BPM orbit data can be used to optimize model parameters, facilitating the identification of quadrupole errors. Furthermore, the fitted results involving accelerating cavities reveal the potential to characterize multipolar fields within the cavities. These findings highlight the versatility and effectiveness of the proposed method. Future work will focus on leveraging the corrected linear model to advance machine learning algorithms for further improvements in accelerator performance.</p>