时间序列数据流的自适应预测

An Adaptive Forecasting Method for Time-Series Data Streams

提出一种自适应预测方法AFStreams,综合了复杂人工智能预测方法和时间序列预测方法的优点,可以根据数据流值变化的快慢程度自适应地确定预测步长,在计算资源受限的前提下,形成最佳预测点轨迹．仿真实验证明,AFStreams能够良好地适应数据的变化,在计算复杂度和预测精度之间平衡,显著地提高了平均预测精度．

An adaptive forecasting method that combines the merits of the precision of artificial intelligence forecasting method and the rapidness of times-series forecasting method, called AFStreams, is proposed. It can estimate the forecasting-step self adaptively from the change ratio of stream-values and can generate proved optimal track of forecasting points with the minimum computation cost from limited resources. Experiments proved that AFStreams can adapt to the changes of data well and provide tradeoff between computing complexity and forecasting precision.