Abstract: In the analysis of rail transit passenger flow, the seasonal characteristics of data have a significant impact on the effectiveness of passenger flow forecast. By analyzing the passenger flow curves of rail transit, it was found that the passenger flow of rail transit presented seasonal characteristics. Aiming at this kind of characteristics, a passenger flow forecast approach of rail transit based on seasonal classification model was proposed. Firstly, the seasonal classification template and seasonal time series were established according to the seasonal characteristics of passenger flow. Secondly, the seasonal classification model of passenger flow was established by the multiplicative seasonal autoregressive differential moving average model. Finally, the seasonal classification model was used to forecast the passenger flow on the corresponding dates. The experiment shows that the proposed seasonal classification model can not only effectively predict the passenger flow of rail transit, but also avoid the fluctuation of prediction error.

Key words: traffic engineering, short term forecast of passenger flow, seasonal classification model, time series, multiplicative seasonal autoregressive differential moving average model

摘要： 轨道交通客流的分析中，数据季节性特征对客流预测的有效性存在显著影响。通过分析轨道交通客流曲线，发现轨道交通客流呈现出季节性特征；针对这种特征，提出基于季节分类模型的轨道交通客流预测方法。根据客流季节特征建立季节分类模板和季节时间序列；采用乘法季节自回归差分滑动平均模型建立客流季节分类模型；使用季节分类模型预测对应类型日期的客流。实验表明：季节分类模型既能有效预测轨道交通客流，又能较好地避免预测误差波动性问题。

关键词: 交通工程, 客流短期预测, 季节分类模型, 时间序列, 乘法季节自回归差分滑动平均模型