地铁新线网站点客流量预测的自训练模型研究

A self-training model for predicting passenger fl ow of stations in new metro networks

准确预测地铁新线网的站点客流量是地铁规划和管理的重要工作。为提高新线网站点客流量的预测精度,文章提出一种迭代创造新样本以扩充训练集的自训练模型。该模型由K最邻近模型和极端梯度提升模型组成。首先,使用K最邻近模型生成与已有训练样本相似的伪标记样本。然后,使用极端梯度提升模型选择最置信的伪标记样本以扩充训练集。在迭代结束后,输入新线网的站点客流影响因素数据,预测站点客流量。结果显示:所提出的模型与几种基准模型相比具有更高的预测精度。

It is the important duty of metro planning and management to accurately predict the passenger flow of stations in new metro networks.In order to improve the prediction accuracy of the passenger fl ow of stations in a new metro network,a self-training model(Self-KNN-XGBoost)is proposed,which iteratively creates new samples to enlarge the training set.The model is composed of the K Nearest-Neighbor model(KNN)and eXtreme Gradient Boosting model(XGBoost).First,the KNN model generates pseudo-labeled samples similar to existing training samples.Then,the XGBoost model is used to select the most confident pseudo-labeled sample to enlarge the training set.After the iteration,the data of influencing factors of station passenger fl ow of new metro networks should be input to predict the station passenger flow.The results show that the proposed model has higher prediction accuracy than several benchmark models.