基于网约车数据的城市区域出行时空特征识别与预测研究

针对交通需求特征识别和需求预测问题，构建改进的LDA(Latent Dirichlet Allocation)城市区域内出行需求识别与预测组合模型，快速识别城市区域内出行需求特征并对需求做出预测. 构建城市交通小区尺度内的空间和时间维度下的主要出行需求特征分布挖掘辨识方法，以及数据集在不同时间尺度下时间维度出行特征构建及预测方法. 利用北京市三环内网约车出行订单数据，验证模型的有效性和准确性. 结果表明，模型能够对不同时间窗口下的区域出行需求特征进行辨识和预测，取得较好的结果.

Regional Travel Demand Mining and Forecasting Using Car-hailing Order Records

This paper proposed a combined model that can fast mining traffic demand and prediction based on the Latent Dirichlet Allocation (LDA) analysis model. This combined analysis framework is able to deal with demand identification and prediction at the same time. The study first developed the traffic demand identification model at the traffic analysis zone (TAZ) scale for presenting the demand characteristics at both the spatial and temporal dimension.Then it proposed a prediction method under a multi-scale time window. The effectiveness and accuracy of the model was verified using car-hailing order data within Beijing's third-ring road. The results show that the model can identify and predict the regional travel demand under different time windows, and achieve good results.