通信学报 ›› 2017, Vol. 38 ›› Issue (12): 98-108.doi: 10.11959/j.issn.1000-436x.2017286

• 学术论文 • 上一篇    下一篇

基于规则前件发生树匹配的数据流预测方法研究

尤涛,李廷峰,杜承烈,钟冬,朱怡安   

  1. 西北工业大学计算机学院,陕西 西安 710129
  • 修回日期:2017-06-11 出版日期:2017-12-01 发布日期:2018-01-19
  • 作者简介:尤涛(1983-),男,河南陕县人,博士,西北工业大学副教授,主要研究方向为分布式数据处理。|李廷峰(1992-),男,黑龙江鸡西人,西北工业大硕士生,主要研究方向为数据库与数据挖掘。|杜承烈(1970-),男,陕西西安人,博士,西北工业大学教授,主要研究方向为军用软件工程。|钟冬(1979-),男,陕西西安人,博士,西北工业大学副教授,主要研究方向为计算机网络技术。|朱怡安(1961-),男,陕西西安人,博士,西北工业大学教授,主要研究方向为并行计算。
  • 基金资助:
    2017航空科学重点基金资助项目;2016复杂产品智能制造系统技术国家重点实验开放研究基金资助项目

Data stream prediction based on rule antecedent occurrence tree matching

Tao YOU,Ting-feng LI,Cheng-lie DU,Dong ZHONG,Yi-an ZHU   

  1. School of Computer Science and Engineering,Northwestern Polytechnical University,Xi’an 710129,China
  • Revised:2017-06-11 Online:2017-12-01 Published:2018-01-19
  • Supported by:
    2017 Aviation Science Key Foundation of China;2016 Open Fund of State Key Laboratory Intelligent Manufacturing System Technology

摘要:

现有基于规则匹配的数据流预测算法存在前件发生定义不准确、前件相关性未考虑、预测结果描述不严谨等不足,造成预测过程效率较低、精度不高等问题。提出基于前件发生树的概率叠加预测算法,定义区间最小非重叠发生,避免前件的错误匹配;通过前件的合并构建前件发生树,提高前件发生的搜索效率;基于概率叠加的思想计算后件的发生区间和发生概率,使预测精度进一步提高。理论分析和实验结果表明,该算法具有较高的时空效率和预测精度。

关键词: 数据流, 情节规则, 区间最小非重叠发生, 前件发生树, 概率叠加预测

Abstract:

There are some shortages in the existing rule-based data stream prediction algorithm,such as inaccurate definition of antecedent occurrence,ignoring the correlation between rules and imprecise description of prediction accuracy.These make low forecasting process efficiency and low prediction accuracy.The superposed prediction algorithm was proposed based on antecedent occurrence tree,and interval minimal non-overlapping occurrence was defined to avoid the problem of excessive matching antecedent.The efficiency was improved for searching antecedent’s occurrence by merging rule’s antecedents in antecedent occurrence tree,and the succedent occurrence based on superposed probability was predicted to enhance prediction accuracy.The theoretical analysis and experimental evaluation demonstrate the algorithm is superior to the existing prediction algorithms in terms of time and space efficiency and prediction accuracy.

Key words: data stream, episode rule, interval minimal non-overlapping occurrence, antecedent occurrence tree, prediction based on superposed probability

中图分类号: 

No Suggested Reading articles found!