Most studies on localization algorithms are conducted on the sensor networks with densely distributed nodes. However, the non-localizable problems are prone to occur in the network with sparsely distributed sensor nodes. To solve this problem, a range-based predictive localization algorithm (RPLA) is proposed in this paper for the wireless sensor networks syncretizing the RFID (WSID) networks. The Gaussian mixture model is established to predict the trajectory of a mobile target. Then, the received signal strength indication is used to reduce the residence area of the target location based on the approximate point-in-triangulation test algorithm. In addition, collaborative localization schemes are introduced to locate the target in the non-localizable situations. Simulation results verify that the RPLA achieves accurate localization for the network with sparsely distributed sensor nodes. The localization accuracy of the RPLA is 48.7% higher than that of the APIT algorithm, 16.8% higher than that of the single Gaussian model-based algorithm and 10.5% higher than that of the Kalman filtering-based algorithm.