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

Streaming data is a collection of dynamic data that grows over time,its data can arrive in real time,and the order of arrival is independent,independent of the control of various application systems;the scale of the data is huge and it cannot be known in advance.The maximum value;once the data is processed,it is not allowed to be taken out again,or it is expensive to extract again,except for intentional saving.The traditional data analysis and streaming data analysis processing are mainly different in the query and processing methods.The traditional data analysis processing is a static process.When the system is not busy,the data is imported into the data warehouse in batches,while the streaming data is continuously injected into the data.In the warehouse,this is a flowing process that processes data dynamically.Through the study of convective data,we can monitor the satellite cloud image,analyze the stock market,judge the network attack,and predict the coal mine disaster.The query and acquisition of events in streaming data is the basis for studying the various operations of streaming data.The time query in the existing streaming data application system only queries the data points of the anomalies in the stream.In the actual situation,the events in the streaming data are mostly anomalies in a continuous time,including time,A variety of information in the spatial location,therefore,the traditional threshold query method can not fully analyze the event from different time and space perspectives,the query accuracy is very low,resulting in the complete failure to obtain all the complete information of the event.In response to these problems,this paper will study the streaming data event acquisition method based on multi-granular Top-k query.Firstly,the whole monitoring area is partitioned,and a projection-based partitioning method is proposed to construct monitoring clusters.Through the monitoring cluster constructed by projection-based partitioning,it can be solved that the monitoring working surface of each monitoring area is often irregular or narrow;and the k-means clustering method is used to select the cluster head nodes in each monitoring area to ensure each area.The selected cluster head node can represent this area;on this basis,the top-k anomaly cluster head nodes of each cluster are monitored according to the severity of the anomaly,and an improved Top-k query is proposed.The method is used to monitor the ordering results of the cluster head nodes,and a regional cross-location method is proposed to obtain the specific location of the cluster head nodes where the anomalous events occur.Secondly,since each cluster head node may have an abnormal event at different time granularities,in order to obtain complete information about the occurrence of an abnormal event,first,events at different time granularities in the cluster head node where the abnormal event occurs most seriously The severity of Tok-k sorting is presented.This paper proposes a sorting method based on chain tree for anomalous events.Then,based on the sliding average method,the query range is expanded in both directions,and the specific range of abnormal data is further determined to prepare for the subsequent construction of the abnormal template.Finally,the experimental analysis of the streaming data event acquisition method based on multi-granular Top-k query is carried out,and the experimental method is verified by the traditional partitioning method and the projection-based partitioning method proposed in this paper.The tree's abnormal event ordering is compared with the traditional sorting to verify the efficiency of the chain-based anomaly event sorting.Experiments show that the proposed chain tree-based anomaly event sorting method is faster.