Roadside Units (RSUs) deployment optimization for vehicle-to-road communication is crucial in improving the performance of the vehicular and transportation networks. Current researches on RSU deployment optimization overlooks the inequal supply-demand relationship of vehicular communications caused by unbalanced vehicular density. In addition, the impact of traffic safety risks on RSU optimization is not considered. To cope with these challenges, this paper presents a data-driven and multi-factor RSU deployment strategy, which can be divided into two stages. The first stage adopts a data-driven approach to analyze vehicular density using realistic vehicle trajectory data, which helps identify preliminary deployment positions based on unbalanced vehicular density. In the second stage, an RSU deployment model is constructed that considers RSU deployment costs, geographical environment constraints, road coverage, and traffic safety risks. This model calculates the cumulative Poisson probability of simple and general traffic crashes to evaluate the traffic safety risks within the RSU coverage range. Afterward, this paper proposes an improved genetic algorithm to obtain the optimal position of RSUs within the deployment area. The algorithm employs a new method for population initialization and adopts a linear ranking-based selection mechanism. Finally, we conduct a joint simulation of transportation and communication networks by linking SUMO with OMNET++ through the TraCI interface and Veins framework. The results evaluate and demonstrate the performance of the proposed method in vehicle coverage, traffic safety risk coverage, and average notification time in comparison with the existing schemes.