摘要
基于交通流预测理论,根据共享单车轨迹数据,研究共享单车出行的时空特性规律。首先,针对出行规律所具有的时间序列特性,建立基于LSTM的线性回归预测模型;其次,针对不同地域间出行的自流动性和关联性,在提出的预测模型基础上,将轨迹数据中挖掘的不同距离特征纳入预测模型,兼顾了共享单车系统不同区域间的空间属性。以已有的300万用户出行记录为基础,将该模型应用于北京共享单车出行需求预测,其结果与已有共享单车需求预测模型的结果相比,精度有明显的提高。
Based on the theory of traffic flow prediction, the temporal and spatial characteristics of shared bicycle travel are studied based on shared bicycle trajectory data. Firstly, based on the time series characteristics of travel rules, a linear regression prediction model based on LSTM is established. Secondly, based on the self-flow and correlation of travel between different regions, the trajectory data is mined based on the proposed prediction model. Different distance features are included in the prediction model, taking into account the spatial attributes between different areas of the shared bicycle system. Based on the existing 3 million user travel records, the model is applied to the forecast of shared bicycle travel demand in Beijing. The result is significantly improved compared with the results of the existing shared bicycle demand forecasting model.
作者
李颖宏
马勇
LI Ying-hong;MA Yong
出处
《智能城市》
2019年第5期1-4,共4页
Intelligent City
关键词
循环神经网络
时间序列特性
关联
共享单车
预测
cyclic neural network
time series characteristics
association
shared bicycle
prediction
