摘要
铁路客票业务流量数据是反映系统业务运行状态的重要记录,为加强流量数据异常预警,针对流量数据具有历史规律性及突变性的特点,选用适于解析数据时间序列依赖度高的门控循环单元神经网络模型(GRU),对流量数据实现时序拟合及趋势预测。GRU采用不同时间步长对流量数据进行拟合的结果在整点或半点周期时间步长具有局部最小特征,该特征与铁路售票时刻规则形成的时间序列依赖规律相一致。在相同数据条件下,使用GRU算法与自回归模型等主流预测算法进行拟合准确度对比,结果证明GRU在解析铁路客票业务流量数据依赖方面具备较高的准确性。经过对异常流量数据趋势预测及拟合,在数据异常区间,预测结果与真实数据的拟合近似验证了GRU算法能够为铁路客票业务流量数据异常预警提供可行性策略。
The traffic data of railway passenger ticket service is important record to reflect the operation status of the system.In order to strengthen the early warning of abnormal traffic data,according to the characteristics of historical regularity and mutation of traffic data,the gated recurrent unit neural network model(GRU)which is suitable for analyzing data with high dependence on time series is selected to realize time series fitting and trend prediction.The results of GRU fitting the flow data with different time steps have the local minimum feature in the whole point or half point cycle time,which is consistent with the time series dependence rule formed by the railway ticketing time rule.Under the same data conditions,GRU algorithm and auto regression model and other mainstream prediction algorithms are used to compare the fitting accuracy.The results show that GRU has high accuracy in analyzing the traffic data dependence of railway passenger ticket service.After the trend prediction and fitting analysis of abnormal traffic data,the difference between the predicted results and the real data is obvious in the data abnormal interval,which verifies that GRU algorithm can provide feasible strategies for the early warning of abnormal traffic data of railway passenger ticket service.
作者
谢泽
朱建生
李雯
XIE Ze;ZHU Jian-sheng;LI Wen(China Academy of Railway Sciences,Beijing 100081,China;Institute of Computing Technologies,China Academy of Railway Sciences,Beijing 100081,China)
出处
《计算机技术与发展》
2021年第10期209-214,共6页
Computer Technology and Development
基金
中国国家铁路集团有限公司科技研究开发计划课题(K2019X008)。
关键词
门控循环单元
流量数据
时序拟合
趋势预测
数据预警
gated recurrent unit
network traffic data
time series fitting
trend prediction
data early warning
