基于历史数据的系统隐患预测研究

摘要

In this paper, a platform performance hidden troubles prediction scheme was presented based on the historical performance of the platform itself combined with LOESS fitting, TBRM fitting and DTW algorithm to predict the abnormal fl uctuation of the platform resources correctly. Practice has proved that this scheme runs correctly and effectively and achieves good performance in use.%基于平台自身历史表现，结合LOESS拟合、TBRM拟合以及DTW 3种算法，提出一种平台性能隐患预测方法，达到准确预测平台层面资源波动异常的目的。实践表明，该方法准确有效，在实际使用中取得了良好效果。