In recent years, with the rapid development of Chinas economy and the continuous increase in population, the urbanization process has intensified. This travel issue caused by peak traffic congestion affects a significant portion of urban populations. In order to conduct a thorough analysis, this article primarily focuses on the urban rail transit industrys subway system. In this study, three models - LSTM, ARIMA and ARIMA-LSTM were employed to analyze and predict short-term inbound flow data for the Hangzhou subway in January 2019. While the LSTM model used for forecasting, it was found to be ineffective in predicting data with extreme peaks. The ARIMA model is subsequently employed for data prediction, revealing its inadequacy in accurately forecasting unstable and non-patterned data. To overcome this limitation, a combined ARIMA-LSTM model is proposed to mitigate the shortcomings of individual models and achieve superior performance. The implementation of the ARIMA-LSTM model effectively mitigates subway congestion issues, thereby facilitating informed decision-making by subway operators. Moreover, it has a certain degree of robustness, even if the other data are not regular enough. The model can be applied in the real subway passenger flow prediction, which is conducive to the choice of people traveling to work and the management decisions of subway operators.