[Purposes] This study conducts interval prediction research on passenger flow at bus stops to address the drawback that traditional point prediction models fail to effectively characterize uncertainty information in bus stop passenger flow forecasting.[Methods]An improved particle swarm optimization (IPSO)-optimized gated recurrent unit (GRU) model combined with kernel density estimation (KDE) is adopted for interval prediction. First, the GRU model is applied to implement point prediction, and the IPSO algorithm is utilized for parameter optimization. Second, kernel density estimation is employed to calculate the probability density distribution function of point prediction errors, so as to obtain error confidence intervals under corresponding confidence levels. Finally, interval prediction is realized by integrating the point prediction results.[Findings]Comparative experiments show that the point prediction module of the proposed model achieves the highest prediction accuracy, and its interval prediction module obtains the optimal interval evaluation metrics, which verifies the validity of the model.[Conclusions] The model can generate more accurate prediction intervals. It can provide references for the decision-making of subsequent bus scheduling schemes and possesses favorable practical value.