License-plate recognition (LPR) data are emerging data sources in urban transportation systems which contain rich information. Large-scale LPR systems have seen rapid development in many parts of the world. However, limited by privacy considerations, LPR data are seldom available to the research community, which lead to huge research gap in data-driven applications. In this study, we propose a complete solution using LPR data for link-based traffic state estimation and prediction for arterial networks. The proposed integrative data-driven framework provides the inference of both cycle maximum queue length states and average travel times of links using LPR data from a subset of intersections in an arterial network. The framework contains three novel data-driven sub-components that are highly customized based on the characteristics of LPR data, including: a traffic signal timing inference model to find signal timing information from the LPR timestamp sequences; a light-weighted queue length approximation model to estimate lane-based cycle maximum queue lengths and a network-wide traffic state inference model to perform network-level estimation and prediction using partially observed data. This study exploits and utilizes the unique features of LPR data and other similar vehicle re-identification data for urban network-wide link-based traffic state estimation and prediction. A six days’ LPR dataset from a small road network in the city of Langfang in China and a more comprehensive link-level field experiment dataset are used to validate the model. Numerical results show that the framework provides good estimation and prediction accuracy. The proposed framework is efficient and calibration-free, which can be easily implemented in urban networks for various real-time traffic monitoring and control applications.