In recent years, with the rapid development of economy and the explosive growth of urban population, China’s traffic congestion has increased. Especially in some provincial capitals, the traffic congestion is still continuing in the peak hours, which not only hinders people’s travel, but also brings many problems to the management of the relevant authorities. Therefore, optimal design of urban transportation planning is an inevitable choice for the sustainable development of the city This paper mainly studies the optimization design of urban traffic planning based on big data, and establishes the traffic signal quality prediction model by using the improved regression tree algorithm. At the same time, according to the spatial distribution pattern and coverage, K-Centroids distributed clustering algorithm is used to adapt to the current situation of base station optimal deployment. The experimental analysis shows that the iterative termination error of dynamic clustering is set to 0.0001, and the optimal projection direction vector is (0.0163, 0.0098, 0.617, 0.5123, 0.6701). The results show that for the problem of unbalanced distribution of training sample data set, smote method is used to reconstruct the sample data set to achieve the balance, and the experimental data also verify the effectiveness of the designed method.