With continuous growth of urban populations, transportation system faces many challenges related to increasing demand of real time services, limited government investment and sustainability of environment. Ride sharing, as a mobile-internet-based transportation service mode, has gained wide popularity across the world. Instead of operating a fixed fleet, a ride sharing platform consolidates supplies from independent drivers with dynamic and flexible schedules. A ride sharing system provides drivers with a flexible working method and also improves passengers’ trip experiences in respect of easy reservation and convenient access to trip information. Unlike traditional taxi business, where supply is constant, ride sharing systems interact with a dynamic fleet. Therefore, adequate study for drivers’ behaviors is of great research interest, as it not only enriches behavioral study for suppliers in economic activities but also supports design of operation strategy for ride sharing system. This research proposes a comprehensive and data driven method that implements behavioral study based on Multinomial Logit Model (MNL) and Mixed Logit Model, from the family of Random Utility Maximization models. Furthermore, in order to explore operation strategies, a simulation framework of ride sharing system is developed. Operation strategies that involve consideration of drivers’ behaviors are proposed and simulated, which attempt to improve the ride sharing system’s operation efficiency.