基于Fisher聚类的公交客流量时间序列预测及对比

Prediction and comparison of bus passenger flow time series based on Fisher cluster

公交客流量预测是城市公共交通管理的基础,科学的客流量预测能够为公交系统管理和路线调整提供可靠依据。考虑到公交客流量的波动差异性以及预测的复杂性,首先利用Fisher算法对原始数据聚类,并依时段划分为六种类型;然后选择自回归差分移动平均模型以及季节性自回归差分移动平均模型两种方法开展公交客流量的预测,并以广州市公交客流量数据进行实证分析,最后计算两种模型的平均绝对误差和平均绝对百分比误差,对比分析基于聚类数据的两模型预测效果的优劣。结果发现:基于Fisher聚类数据,季节性自回归差分移动平均模型的预测效果较好,且比数据未聚类前对应模型预测的效果更优。

The prediction of bus passenger flow is the basis of urban public transport management.Scientific prediction of passenger flow can provide a reliable basis for the management of bus systems and route adjustment.Considering the dissimilarity of fluctuation and the complexity of prediction of bus passenger flow,the original data was firstly clustered by using Fisher’s algorithm and was divided into six types according to time interval.Then,the autoregressive differential moving average model and seasonal autoregressive differential moving average model were selected to make predictions of public traffic passenger flow,and empirical analysis was conducted based on Guangzhou public traffic flow data.Finally,the average absolute errors and the average absolute percentage errors of the two models were calculated,and the prediction effect of the two models based on clustering data were compared and analyzed.The results show that based on the Fisher clustering data,the seasonal autoregressive differential moving average model has a better prediction effect,which is much better than that with data without clustering.