基于粒子群优化-支持向量回归的高速公路短时交通流预测

为实现高速公路短时非线性交通流的精准预测,依托高速公路运营积累的大量数据资源,构建了基于粒子群优化(par-ticle swarm optimization,PSO)的支持向量回归(support vector regression,SVR)预测模型.首先,对获取的高速公路交通流数据进行异常值剔除、缺失值填充以及归一化等预处理;其次,基于SVR算法采用滑动窗口的方式建立预测模型,并基于具有较强寻优能力的PSO优化算法获取SVR模型的最优参数组合;最后,通过京台高速济南西收费站断面交通流数据进行实例验证.模型的预测结果表明,所提出的高速公路短时交通流预测模型能够满足实际需求,且相较反向传播(back propagation,BP)、差分整合移动平均自回归模型(autoregressive integrated moving average model,ARIMA)模型具有较高的准确性,可为日后高速公路运营决策提供理论支持.

Short-term Traffic Flow Prediction of Expressway Based on PSO-SVR

In order to realize the accurate prediction of the short-term nonlinear traffic flow of the expressway, a Support Vector Regression (SVR) forecast model based on Particle Swarm Optimization (PSO) was constructed. It relys on the large amount of data resources accumulated by the expressway operation. Firstly, the expressway traffic flow data was preprocessed, such as outlier removal, missing value filling and normalization; secondly, based on the SVR algorithm, the prediction model was established by means of sliding window, and the optimal parameter combination of the SVR model was obtained based on the PSO optimization algorithm with strong optimization ability; finally, the cross-section traffic flow data of the Jinan West Toll Station of Shandong Expressway was used for instance verification. The prediction results of the model show that the proposed highway short-term traffic flow prediction model can meet actual needs. It has higher accuracy than the BP and ARIMA algorithms, and can provide theoretical support for future highway operation decisions.