基于多粒度Top-k查询的流式数据事件获取方法

Streaming Data Event Acquisition Method Based on Multi-granular Top-k Query

流式数据中事件的查询及获取是研究流式数据各类操作的基础.现有流式数据系统中的事件查询只针对流中的异常数据点进行查询,而在实际情况下,流式数据中的事件多为一段连续时间的异常,包含时间、空间位置等多种信息,因此,传统的阈值查询方法无法从不同的时间及空间角度对事件进行全面分析,查询准确性极低,导致无法获取事件的全部信息.针对这些问题,本文提出一种基于多粒度Top-k查询的流式数据事件获取方法.该方法首先在监测区域内构建基于投影分区的区域监测簇;在此基础上,提出一种改进的多粒度空间Top-k查询方法对点进行查询,获取异常事件发生的空间位置信息;其次,基于事件峰谷点信息,对流式数据进行时间多粒度Top-k查询,找出异常事件的触发点和终止点,从而获得事件的完整信息.实验表明,本文提出的方法在系统资源开销、查询效率等方面均具有很大优势.

The acquisition of events in streaming data is the basis for studying various types of query processing of streaming data.For the current streaming data system,the specific location information of the event cannot be quickly determined in the monitoring area,and the event acquisition efficiency and accuracy are low.The problem in this paper is to propose a streaming data event acquisition method based on multi-granular Top-k query.Firstly,the regional monitoring cluster based on projection partition is constructed in the monitoring area.On this basis,an improved multi-granular Top-k query method is proposed.The abnormal cluster head point is queried to obtain the spatial location information of the abnormal event;secondly,based on the event peak and valley information,the time-multi-granular Top-k query is performed on the streaming data,and a streaming data event based on sliding window is proposed.The get method obtains the complete information of the event by finding the trigger point and the end point of the abnormal event in the window.Experiments show that the method has great advantages in system resource overhead and query efficiency.