基于Softmax函数增强卷积神经网络-双向长短期记忆网络框架的交通拥堵预测算法

对交通状态进行预测，就需要准确识别和判断交通状态。该文没有采用传统的以车辆速度为基准的预测方法，而是使用TTI交通拥堵系数，该系数的计算基于道路自身的自由流速度，可以让具有不同速度等级的街道都统一到TTI系数上来作为拥堵评价，因此相较以传统的车辆速度为基准的预测方法更能表现出道路的拥堵状态。该文提出了一种改进的深度学习预测模型（CS-BiLSTM），该模型基于卷积神经网络（CNN）和双向长短期记忆网络（BiLSTM），并结合Softmax函数增强CNN提取出的交通空间特征信息。深度学习预测模型（CS-BiLSTM）中的S代表的就是Softmax的缩写。使用成都市出租车GPS数据进行验证，结果表明，所提出的CS-BiLSTM模型具有更高的准确性，其性能相比C-BiLSTM网络预测框架提升了13%。

Research on Traffic Congestion Prediction Algorithm Based on CS-BILSTM Framework

In order to predict the traffic state, it is necessary to identify and judge the traffic state accurately. Instead of using the traditional prediction method based on vehicle speed, the TTI traffic congestion coefficient is used by this paper, which is calculated based on the free flow speed of the road itself, so that streets with different speed levels can be unified into the TTI coefficient as the congestion evaluation.Therefore, compared with the prediction method based on the traditional vehicle speed, it can better show the road congestion state. An improved deep learning prediction model (CS-BiLSTM) is proposed by this article, which is based on convolutional neural network (CNN) and bidirectional long short-term memory network (BiLSTM), combined with Softmax function to enhance the traffic spatial features extracted by CNN information. The S in the Deep Learning Predictive Model (CS-BiLSTM) stands for the abbreviation of Softmax. Based on taxi GPS data in Chengdu, the results show that the proposed CS-BILSTM model has higher accuracy, and its performance is improved by 13% compared with the C-BILSTM network prediction framework.