Electric vehicles (EV) face great challenges, i.e., short driving range, long charging time, and few charging stations. Especially, these characteristics of EV hamper its acceptability to taxi drivers. Leveraging massive taxi GPS trajectories, this study presents a novel spatial-temporal route recommendation system for e-taxi driver by integrating spatial-temporal data mining and optimization. Taxi travel knowledge, including the probability of picking-up passengers and the distribution of client destination, is learned from raw fuel taxi GPS trajectories. Considering the cascading effect of route decision-making, consecutive actions of an individual electric taxi (ET) driver is modeled by an action tree. The expected net revenue of ET driver’s actions is estimated with the learned knowledge by incorporating cruising routes and charging decisions. An online recommendation prototype system is developed for high-efficient real-time route recommendations, i.e., going to a charging station, cruising on some roads, etc. An experiment using real-world GPS trajectories of 16,146 fuel taxies is conducted on the performance evaluation. Results show that the net revenue of the ET drivers is up to 76.2% better than real-world fuel taxi drivers. The presented approach not only increases the revenue of ET drivers but also improves the EV viability.