This paper considers a ridesharing problem on how to match riders to drivers and how to choose the best routes for vehicles. Unlike the others in the literature, we are concerned with the maximization of the average loading ratio of the entire system. Moreover, we develop a flow-dependent version of the model to characterize the impact of pick-up and drop-off congestion. In another extended model we take into account the riders’ individual evaluation on different transportation modes. Due to the large size of the resulting models, we develop a large neighbourhood search algorithm and demonstrate its efficiency.