This paper proposes a dynamic fingerprinting combination (DFC) algorithm that improves mobile localization by dynamically weighting the spatial correlation from multiple location fingerprinting systems. This DFC algorithm first extracts the complementary advantages of fingerprinting functions to construct a fusion profile, and then dynamically combines individual outputs based on the fusion profile surrounding the test sample. The proposed algorithm generates more accurate location estimates and reduces the risk of selecting a poorly-performing fingerprinting approach. This study applies DFC to an actual GSM network with realistic measurements. Experimental results show that DFC improves the positioning accuracy of base fingerprinting algorithms, including the Bayesian approach and a neural network model.