摘要： 研究了应用数据挖掘技术预测时间序列数据中事件的方法。针对时间序列数据提出了显著特征提取算法，给出了特征间的相似度量标准，并应用特征聚类算法，将时间序列数据转换成相应的特征序列表示。应用频繁模式发现算法和预测模式生成算法在预测时段内发现与目标事件相关的时序特征模式，预测事件的发生。实验结果表明，该文所提出的方法能够有效地预测时间序列数据中的事件。

关键词: 事件；时间序列；数据挖掘；特征；聚类

Abstract: A research on event prediction in time series based on data mining method is undertaken. A significant feature extraction algorithm and a corresponding feature similar measure are proposed. Time series is transformed into feature sequence by application of the clustering algorithm. Prediction pattern for event is extracted in prediction period through using frequent feature pattern searching algorithm and prediction pattern generating algorithm. Experimental results indicate the proposed method is effective in event prediction

Key words: Event; Time series; Data mining; Feature; Clustering