为准确预测地铁客流量，以便更好地进行车辆调度和站点管理，提出一种基于 K 近邻模式匹配的地铁客流量短时预测方法． 通过分析地铁客流数据，发现地铁的日客流发展模式具有一定规律;针对该发现，提出一种基于误差变化率计算的自适应 K 值获取算法，自动获取合适的 K 值以提高预测算法的普适性． 最后以广州火车南站地铁站为例进行实例分析，结果表明:所提出的方法同时适用于节假日与非节假日两种不同交通模式的地铁客流预测，平均预测精度在 90%左右，具有较好的应用推广价值．

In order to forecast the subway traffic accurately and better the vehicle scheduling and site management，a short-term forecasting of subway traffic based on K-nearest neighbour pattern matching is proposed． By analyzing the subway traffic data，it is found that the day passenger flow development mode of subway has a certain law．Thus，an adaptive K acquisition algorithm based on the calculation of error rate is proposed，it can improve the uni-versality of the prediction algorithm by automatically obtaining the appropriate K． Finally，Guangzhou South Rail-way Station is taken as an example for the case study． The experimental results show that the method proposed is applicable to the subway passenger flow forecast of two different modes of traffic-holidays and non-holidays，its ave-rage prediction accuracy is about 90%． It can be seen that this method has a good application value．