Abstract: We aim to overcome the problem of noise sensitivity of high-frequency GNSS data in source mechanism solution inversion. We proposes a Correlation-based Complete Ensemble Empirical Mode Decomposition (CEEMD)-Wavelet denoising and multi-source fusion inversion framework, aiming to enhance high-frequency GNSS positioning accuracy and thereby improve the precision of source mechanism inversion. Taking the 2018 Anchorage, Alaska, America Mw 7.0 earthquake as a case study, the variance reduction (VR) values for raw high-frequency GNSS, denoised high-frequency GNSS, strong-motion seismograph, and fused data reached 67.2%, 76.0%, 80.1%, and 95.1% respectively. The denoising process improved waveform fitting accuracy by approximately 10%, while multi-source data fusion achieved a 28% enhancement. Comparative analysis of source parameters (strike/dip/slip angles, magnitude, depth) with GCMT and USGS reference solutions confirmed the consistency of fusion-based inversion results. This integrated framework not only significantly enhances the signal-to-noise ratio of high-frequency GNSS data and the reliability of source mechanism solutions, but also provides a multi-source data fusion methodology for regional stress field analysis. It demonstrates substantial practical value for optimizing earthquake early warning systems and investigating fault motion dynamics.