基于Dropconnect的CorrelationNet 预测乘客出租车需求

Passenger taxi boarding demand prediction via CorrelationNet with Dropconnect

为了提高乘客出租车需求预测的准确性以及出租车服务的运行效率,提出一种使用Dropconnect方法的CorrelationNet。该方法包括两个阶段,时空特征选择和Dropconnec正则化。首先,预测模型在深度神经网络中添加了时空相关性分析机制,并确定了用于出租车需求预测的时空特征,形成了新的深度学习网络CorrelationNet。然后,预测模型使用Dropconnect训练新的深度学习网络CorrelationNet,以减少过拟合。最后,在广州进行了应用试验以验证模型;采用支持向量机回归(SVR)、人工神经网络(ANN)和CorrelationNet方法,用相同的出租车GPS数据集来评估所提出方法的性能。实验结果表明,该方法优于其它几种方法,更适合于乘客出租车需求预测。

To improve the accuracy of taxi customer pick-up prediction and the operation efficiency of taxi service,a CorrelationNet with dropconnect method is proposed.The method includes two stages,spatiotemporal features identification and regularization with dropconnect.First,the prediction model adds the spatiotemporal correlation analysis mechanism to the deep neural network,and determines the spatiotemporal features for taxi demand prediction,forming a new deep learning network called CorrelationNet.Then,the prediction model uses dropconnect method to train the new deep learning network Correlationnet to reduce over fitting.Finally,a case study is carried out in Guangzhou to verify the model;support vector machine regression(SVR),artificial neural network(ANN)and CorrelationNet methods are adpoted to evaluate the performance of the proposed method by using the same taxi GPS data.The experiments results show that this method is better than other methods and is more suitable for passenger taxi demand prediction.