Abstract: The majority of real-world networks are directed, weighted and dynamic. Aiming at the problem of node role analysis in directed weighted network, a novel node role analysis method based on directed topological potential is proposed, which can divide nodes into four roles based on their behavior pattern and local influence. Then, a node role-based dynamic networks evolution analysis method is introduced, which can model the dynamics of behavioral roles representing the main connectivity patterns over time and detect the significant event. The experiment results indicate that proposed approaches can effectively estimate the node role and detect the dynamics of network evolution.