基于支持向量回归的地铁进站客流短时预测模型

基于准确的未来客流信息对地铁运营的重要性,研究客流预测的方法。选取支持向量机应用领域的一大分支——支持向量回归的方法对地铁进站客流进行短时预测,使用一种改进的粒子群算法进行参数寻优,从而构建客流预测模型。提出的模型以日期类型和所处时刻作为输入,可以提前预测未来一周的每15 min的客流。采取平均绝对百分比误差和均方根误差对模型的预测结果进行评估。使用广州杨箕车站进站客流数据进行实验,通过交叉验证确定验证参数选取的合理性,并将该模型与BP神经网络、KNN算法进行比较,实验表明模型预测结果的精度更高,稳定性更好。

Short-term Prediction Model of Subway Entry Passenger Flow Based on Support Vector Regression

A method of passenger flow prediction was studied because of the importance of accurate future passenger flow information to subway operations. A large branch of the application of a support vector machine, support vector regression, was selected for short-term prediction of subway entry passenger flow. By using an improved basic particle swarm optimization algorithm for parameter optimization, a passenger flow prediction model was constructed. The model proposed takes the date type and the time of the moment as input and can predict the passenger flow for every 15 min. The mean absolute percentage error and root mean square error were applied to evaluate the model''s predictions. Experiments based on passenger flow data from Guangzhou Yangji Station were carried out, and the rationality of parameter selection was determined and verified by cross validation. Compared with the backpropagation neural network and k-nearest-neighbors algorithm, the proposed model has higher accuracy and better stability.