Abstract: A short-term urban rail transit flow prediction method is established to minimize the nonlinearity and randomness of urban rail transit flow in a short period. The urban rail transit stations can be clustered to four clusters using the K-means method, which are urban rail transit interchange station, multimode transportation terminal, commuting station, and common station. The ARMA model and the wavelet ARMA combination model are then applied to predict the short-term urban transit flow using both trend and random errors features of the original flow to predict the flow in the next time interval. The prediction results using field-collected data from Xi’an metro line 2 indicate that the wavelet ARMA combined model predictive effect is better than the single ARMA model, and Wavelet ARMA combination model can predict the flow for each category stations with reliable accuracy and robustness. The proposed method can provide a theoretical reference for urban rail transit operation management. Keywords: urban rail transit, clustering analysis of metro station, short-term prediction, wavelet transform, auto regressive moving average model

CICTP