改进支持向量回归机的短时交通流预测

短时交通流预测是实施智能交通控制的基础和保障.针对目前短时交通流预测方法拟合交通数据的能力偏弱，以及过分依赖历史数据的不足，提出一种基于深度学习回归机的短时交通流预测方法.首先构建深度学习回归机算法模型，包括受限玻尔兹曼机的显层节点输入端，受限玻尔兹曼机的若干中间层，以及径向基支持向量回归机输出端.通过实验将深度学习回归机预测方法与其他典型的短时交通流预测算法进行比较，结果表明，在相同的数据和计算平台下，本文提出的深度学习回归机预测方法精度更高，且预测实时性也能满足实际的需求.

Promoted Short-term Traffic Flow Prediction Model Based on Deep Learning and Support Vector Regression

Short- term traffic flow prediction is the basis of an Intelligent Transport Systems (ITS) project. However, in current practice, the methods for short-term traffic flow prediction have encountered many challenges in fitting the traffic flow data, one is it depends too much on historical data. Therefore, a novel short-term traffic flow forecasting method based on Deep Learning and Support Vector Regression (DL-SVR) is proposed in this paper. Firstly, the DL-SVR model is composed by a Restricted Boltzmann Machine (RBM) visible inputting layer with some RBM intermediate layers and a radial SVR output layer. Furthermore, in order to enhance the generalization of the model, an improved Particle Swarm Optimization (PSO) algorithm is designed to optimize the number of nodes in the inputting layer. Finally, the DL-SVR method is compared with other typical short-term traffic flow prediction algorithms on the same computing platform. The experimental results show that the proposed DL-SVR method gets a higher accuracy in its real-time prediction.