Cold-atom absolute gravimeters are widely used for measuring the acceleration of gravity, yet their sensitivity is often limited by ground vibrations. Existing vibration compensation algorithms struggle to strike a balance between search accuracy and computational efficiency and are prone to local optima. Here, we propose an improved whale optimization algorithm (IWOA) to address these issues. By combining Logistic-LHS (Latin hypercube sampling) chaotic initialization, adaptive adjustment, and a Gaussian mutation operator to prevent premature convergence, IWOA achieves higher search efficiency and superior sensitivity than traditional algorithms. The method is validated through multiple simulation studies and further assessed experimentally on the NIM-AGRb-1 cold-atom gravimeter system. The results show that IWOA reduces the uncertainty of the fitted phase parameter by 66%. The Pearson correlation between atomic transition probability and the calculated phase increases to a maximum of 0.98, and the gravity sensitivity improves to 47 $\mathrm{\mu }\mathrm{Gal}/\sqrt{\mathrm{Hz}}$ when the evolution time T is 80 ms.