With the development of society, people's travel demand is increasing. Traffic congestion has become an urgent problem to be solved. The effective prediction of passenger flow has become a key problem to solve the traffic congestion problem. There is always an accurate prediction result that allows people to better deal with the future traffic situation and make the best response strategy. In this study, the ARIMA model was used for prediction experiments. By analyzing and modeling the ridership data of New York taxi companies from January to July 2015. Get the suitable model for this study is ARIMA (3,1,1) model. And whether the data can be used by ARIMA model is tested to ensure the reliability of the model. Then, the line chart and prediction data about time and number of trips are obtained. Through this study, it can be found that ARIMA prediction model has certain practicability and reliability in predicting data. Through the analysis of the data set, the predicted value of the passenger flow from August 1 to 8 can be obtained more accurately. At the same time, there are some limitations, that can be improved when combined with other models. For example, when dealing with the data set related to vehicles in this study, the ARIMA model could not be used for predictive analysis because the data set itself had many extreme values and outliers.