摘要： 提出一种基于异常值检测的电梯交通流递归预测方法。对电梯交通流进行时间序列分析得到初始季节时间序列模型，引入异常值检测过程，检测出训练数据中的异常值并进行修正，利用修正序列得到最终的季节时间序列模型。把最终的季节时间序列模型转化为状态空间形式，通过卡尔曼滤波实时调整状态向量，实现电梯交通流的在线预测。仿真结果证明该方法有效。

关键词: 电梯交通流预测, 季节时间序列模型, 异常值检测

Abstract: This paper proposes a recursive forecast method for elevator traffic flow based on outlier detection. It analyzes elevator traffic flow data using seasonal time series and gets a initial model, then starts the outlier detection module and gets corrected series. It can get the finial SARIMA model of elevator traffic flow. It transforms the finial SARIMA model to state space model, adjusts the state vector using Kalman filter, and realizes the on-line forecast. Result of simulation shows that this method is valid.

Key words: elevator traffic flow forecast, SARIMA model, outlier detection

中图分类号: 

• TP18