基于EMD-KNN-BP组合模型的高速公路短时流量预测

Expressway Short-Term Flow Prediction Based on EMD-KNN-BP Combined Model

摘要: 为了高效、准确地预测高速公路短时交通流量，本文建立EMD-KNN-BP的组合模型进对高速公路短时流量进行预测，并以山东省高速公路某收费站数据进行实证分析。研究发现：1) 基于EMD分解的组合模型能够提高短时交通流预测的精度，EMD-BP、EMD-KNN相较于单一模型的MSE、RMSE分别降低了24.4%、13.4%和9.6%、4.9%；2) 根据IMF分量特征的不同建立的EMD-KNN-BP模型有效地降低了预测误差，使预测效果达到最优，MSE、RMSE的误差分别降低了61.8%、38.2%。

Abstract: In order to predict expressway short-term traffic flow efficiently and accurately, this paper establishes the combination model of EMD-KNN-BP to forecast expressway short-term traffic flow, and makes an empirical analysis with the data of a expressway toll station in Shandong Province. The results show that: 1) The combined model based on EMD decomposition can improve the accuracy of short-term traffic flow prediction. Compared with MSE and RMSE of single model, EMD-BP and EMD-KNN decrease by 24.4%, 13.4%, 9.6% and 4.9%, respectively. 2) The EMD-KNN-BP model established according to the different characteristics of IMF components effectively reduced the prediction error and achieved the optimal prediction effect. The errors of MSE and RMSE were reduced by 61.8% and 38.2%, respectively.