Association Rule Mining Method For Marine Abnormal Events Based On Time Series Raster

Data Mining and Knowledge Discover(DMKD)is a discipline that extracts valuable information from massive data.It combines lots of frontier discipline such as statistics,computer science,management and so on.With the rapid development of science and technology,human has entered the information age.Data mining and knowledge discover will be fully developed and applied under the background of historical.It should be certained that DMKD is an effective method for erxtracting information from huge data at present.In recent years,with the interpenetrative degree between different disciplines is deepened,and interdisciplinary researchbetween DMKD and other disciplines has received wide attention in academic and commercial field.The study of geography event model and geography event mining has been fully developed in disciplineof geography.A large number of geography event models are proposed,which enrich the research area greatly.Different geography event algorithms were proposed by a great many of exports based on different research direction.Marine events mining is an important research branch of geography event mining.Because of the importance of marine on climatic change and environmental change,it has a science significance and practical significance to study marine event mining for rigional air-sea interaction and global climate change.After studying and analyzing geography events and association rule mining,this paper proposed a marine abnormal event-oriented association rule mining algorithm for continuous ocean phenomenon,and the Pacific Ocean,which has active marine environment factors,is selceted as a research area.The main workof this paperis as follows.(1)On the basis of a mass of references,we detailedly introduced research status of association rule mining in data mining and geography event,which included geography event model domain and geography event mining domain.(2)We expounded detail construction process of event-oriented spatial temporal transaction table.First we detailedly introducedsome concepts and definitions related to events based on marineenvironment,then extraction process of marine factors abnormal state was expounded,including the extraction of monthly anomaly,discretization of marine factors and extraction of marine factors abnormal state at one moment.Finally,we constructed spatial temporal transaction table based on above and cleaned the noise from the table.(3)Association rule mining algorithms for marine abnormal events are proposed.This paper first introduced the core idea of Apriori algorithm,namely found all frequent itemsets by setting support and generated strong association rules from the frequent itemsets by setting confidence.On this basis,we proposed two association rule mining algorithms which dealing with marine abnormal events with the background of marine environment,the first one is CE-ARMA algorithm aiming at co-occurrence marine events;the other one is MAETP-ARMA algorithm aiming at non-simulaneous marine events.Although characteristic of these two algorithms is the same as “link-prune” of Apriori algorithm,they addressed the problem that Apriori prototype algorithm cannot deal with continuous marine abnormal events.Before we introuduce these two algorithms,we firstly introduce some concepts and definitions related to algorithms,then we used a example to expound the operational process of the algorithms.(4)Because of the close association relationship between global climate change and regional sea-air interaction,We used marine environmental parameters such as sea surface temperature,sea level abnormal and so on as our test data and selected Pacific Ocean as our main research area.Finally,we analyzed these two algorithms from computation complexity and algorithm efficiency.