<p>Geomagnetic data inversion is applied for fast and accurate parameter estimation associated with ore and mineral exploration. The inversion is classified as a composite and nonlinear problem, making it a difficult task. Most existing algorithms involve time-consuming parameter tuning. The original JAYA algorithm is a global optimization method that does not require parameter tuning; however, solutions frequently become stuck at local minima. Its solution update relies solely on the best and worst individuals, which increases the risk of convergence to a local minimum. A modified JAYA algorithm (MJAYA) was proposed to overcome this limitation. MJAYA introduces four operators to balance exploration and exploitation, with three learning strategies to improve exploration capability and one to enhance exploitation capability. A rank-based approach is used to support operator selection. MJAYA performance was demonstrated utilizing synthetic geomagnetic data (noise-free and noise-added) and four field geomagnetic anomalies recorded from mineral exploration sites in Sweden, Canada, and India. The MJAYA was compared to five JAYA variant algorithms and outperformed them in convergence, robustness, and solution stability. It achieved the lowest objective function, which was 3.3–878% lower than those from other JAYA variants in both synthetic and field data inversions. Model parameters estimated from field data were consistent with published literature, geological information, and drilling data. In addition, principal component analysis was applied to reconstruct the cost function topography for uncertainty appraisal. The results demonstrate that the MJAYA provides reliable parameter estimations within a defined confidence interval.</p>