时间序列数据流预测模型应用研究

Application research for prediction of time series data stream

时间序列数据流中蕴含了大量潜在信息,可以作为智能决策的依据。研究时间序列数据流的变化趋势,预测其未来一段时间的可能值,能够为当前的决策提供重要的支持。提出用链式可重写窗口技术代替传统的滑动窗口技术,并结合经验模式分解和径向基神经网络建立时间序列数据流在线预测模型——Online_DSPM。实验结果表明,与单一时间序列数据流预测模型相比,该模型具有较高的预测精度和校好的模型适应性。

Time series data stream contains a large amount of potential information that can be used as the basis for intelli- gent decision-making.It can provide an important support for the application of real-time decision by researching data stream prediction.Therefore,re-writable linked window technology is proposed that can replace the traditional sliding window technol- ogy, and combined with empirical mode decomposition and radial basis neural networks one online time series data stream prediction model is established called Online DSPM.The experimental results indicate that the combined model has higher precision of prediction and better adaptability,compared with other single time series prediction models.