In these years, wireless applications based on subscriber location information have been receiving more attention. One of the most popular research areas is radio location techniques of wireless cellular networks. RBF neural network (Radial Basis Function Network) is a typical supervised learning feedforward neural network. Taylor-series expansion location algorithm based on the RBF neural network was proposed in some papers. In this paper, an adjustable localization algorithm with more accurate and lower cost is proposed. The simulation results indicate that this algorithm can obtain different RMSE performance with decision threshold E. When threshold E is low value, the performance is better than Chan's algorithm, LS's algorithm, Taylor's algorithm and LS-Taylor's algorithm in NLOS environment.