基于时间序列的滴滴出行生成预测

Didi Trip Generation Prediction Based on Time Series

摘要:

为了更有效地预测滴滴出行生成，进而使得出行生成预测结果能够更好地运用在交通需求预测的各个方面，在传统交通调查的基础上，针对如何基于时间序列方法实现滴滴出行生成预测进行研究。通过对滴滴打车数据的处理，分析滴滴出行的时间和空间特性，研究滴滴出行的时空变化规律。结合具体案例，提出一种基于ARMA模型的滴滴出行生成预测方法，先计算时间序列的自相关性和偏相关性，确定模型阶数，最终得到出行生成预测值。预测结果表明，基于时间序列方法的滴滴出行生成预测方法能够实现滴滴出行时空变化的短期预测，并取得较好的预测精度，能够为交通规划、交通管理等政策的制定提供经验和参考。

Abstract:

In order to predict the generation of Didi Trips more effectively, the trip generation prediction results can be better applied to all aspects of traffic demand prediction. Based on the traditional traffic survey, how to implement the Didi trip generation prediction based on the time series method is studied. Through the processing of Didi taxi data, the time and space characteristics of Didi Trips are analyzed, and the spatio-temporal variation rules of Didi Trips are studied. Combined with the specific cases, a prediction method of Didi trip generation based on the ARMA model is proposed. The autocorrelation and partial correlation of time series are firstly calculated, and then the model order is determined. Finally, the trip generation prediction value is obtained. The prediction results show that the Didi trip generation prediction method based on time series method can realize the short-term prediction of the spatio-temporal change of the Didi Trips, and the better prediction accuracy is obtained, which can provide experience and reference for the formulation of traffic planning, traffic management and other policies.