摘要： 为解决不确定数据流的预测问题，根据数据流高速、无限和动态不确定性的特点，在复杂人工智能预测和时间序列预测的基础上，提出一种基于优化策略的预测方法。综合考虑数据流中元组的不确定性与不确定异常性，以降低预测计算代价。同时考虑不确定的统计特性对卡尔曼滤波预测的影响，对Q和R进行异步优化估计，以形成最佳状态预测。实验结果表明，该方法的预测性能较好。

关键词: 时间序列, 不确定数据流, 优化估计, 卡尔曼滤波, 复杂度

Abstract: On account of data stream for high-speed, unlimited and dynamic characteristics of uncertainty, sophisticated artificial intelligence forecasting methods and the rapidness of times-series forecasting method is used. A forecast method foruncertain data streams based on optimal policy that combines data stream tuples uncertainty and uncertainty abnormality for reducing the computational cost of forecast is proposed. Taking into account the statistical properties of the Kalman filteing prediction uncertainty on the impact of Q and R, Q and R are estimated by the innovation based asynchronous adaptive estimated at the same time. Experimental results on actual data source show that this method can adapt to the uncertain of data streams well and provide precise instantaneous detection under certain conditions.

Key words: time series, uncertain data stream, optimal estimation, Kalman Filtering(KF), complexity

中图分类号: 

• TP311