Travel timeprediction is an important components of intelligent transportation systems.Based on detected vehicle travel time data,this paper applied Principle Component Analysis (PCA)-Gradient Boosting Decision Tree (GBDT) method to propose a model for travel time prediction of urban road up to 2 h.PCA algorithm is used for travel time series decomposition and principle component extraction,after which GBDT method is adopted to make travel time prediction.The Results indicate that the proposed PCA-GBDT algorithm proves to outperform the conventional k-Nearest Neighboring (kNN) method,time series ARIMA model and support vector machine (SVM) approach with highest prediction accuracy and enough algorithm stability.%城市道路旅行时间预测是城市智能交通管理系统和交通信息服务系统的重要基础.利用实测的车辆旅行时间数据,提出了进行城市道路旅行时间多步预测的主成分分析-梯度提升决策树(PCA-GBDT)方法.首先使用主成分分析方法进行旅行时间序列分解和主成分提取,之后建立了利用梯度提升决策树方法的旅行时间多时段预测模型.实际案例结果表明,与传统kNN方法、时间序列ARIMA方法、支持向量机(SVM)方法相比,PCA-GBDT方法具有更高的预测精度与算法稳定性.
展开▼