基于融合模型动态权值的短期客流预测方法

针对传统交通系统中短期客流预测精度低的问题，考虑城市交通站点客流数据在横纵向时间序列的规律性，基于卡尔曼滤波算法和K近邻（K-Nearest Neighbor, ANN）算法，分别根据当日数据和历史数据对客流量进行预测，然后利用权重系数方程对两个预测值加以融合，从而构建基于融合模型动态权值的短期客流预测方法。以某城市的某公交站点客流数据为研究对象，对所建融合模型短期客流预测的准确性和适用性加以验证。结果表明，新建模型、单一的卡尔曼滤波模型和KNN模型的平均相对误差分别为3.6%, 9.0%和7.7%，可见新建模型能更好地拟合客流变化趋势且评价效率更高。

Short-Term Passenger Flow Prediction Method Based on Dynamic Weight of Fusion Model

In order to solve the problem of low accuracy of short-term passenger flow prediction in traditional transportation system, considering the regularity of transverse and longitudinal time series for passenger flow data at urban traffic stations, the short-term passenger flow was predicted according to current data and historical data respectively based on Kalman filter algorithm and K-Nearest Neighbor(KNN) algorithm respectively. By using the dynamic weights coefficient equation to fuse the two predicting values of the Kalman filter algorithm and KNN algorithm, a new short-term passenger flow prediction method based on the fusion model was constructed. Taking the passenger flow data of a bus station in one city as an example, the accuracy and applicability of the proposed fusion model for short-term passenger flow prediction was verified. The results show that the average relative error of the new model, the single Kalman filter model and KNN model is 3.6%, 9.0% and 7.7%. It means that the new model can better fit the trend of passenger flow and has higher efficiency.