基于小波分析和集成学习的短时交通流预测

针对短时交通流具有随机性和不确定性等特征，本文提出一种基于小波分析和集成学习的组合预测模型。首先对原始交通流数据的平均行程时间序列应用Mallat算法进行多尺度小波分解，且对各尺度上分量进行单支重构；然后对于各重构的单支序列分别使用极端梯度提升模型（extreme gradient boosting, XGBoost）进行预测得到多个子模型，同时利用贝叶斯优化算法进行子模型的最佳参数选择；最后把所有子模型的预测值代数求和，得到总体交通流的预测结果。采用美国纽约布鲁克林地区某路段实际交通流数据进行预测，并对预测结果与其他模型进行比较分析。研究结果表明：小波分析和XGBoost组合模型预测效果优于传统线性模型及单一XGBoost模型，从而更好地为交通管理提供指导意见。

Short-term Traffic Flow Forecasting Based on Wavelet Analysis and Ensemble Learning

In view of the randomness and uncertainty of short-term traffic flow, a combined forecasting model based on wavelet analysis and ensemble learning was proposed. Firstly, the average travel time series of the original traffic flow data were decomposed by Mallat algorithm, and the components on each scale were reconstructed by a single branch. Then, for each reconstructed single branch series, the extreme gradient boosting(XGBoost) model was used to predict and obtain multiple sub models, and the Bayesian optimization algorithm was used to select the best parameters of the sub models. Finally, the predicted values of all the sub models were algebraically summed to obtain the predicted results of the overall traffic flow. The actual traffic flow data of a road section in Brooklyn, New York, USA was used to predict, and the prediction results were compared with other models. The results show that the prediction effect of the combined model of wavelet analysis and XGBoost is better than that of the traditional linear model and single XGBoost model, so as to provide better guidance for traffic management.