基于时序栅格的海洋异常事件关联规则挖掘方法研究

发布时间：2018-02-20 15:51

本文关键词： 关联规则事件挖掘算法海洋异常事件太平洋　出处：《重庆交通大学》2016年硕士论文　论文类型：学位论文

【摘要】：数据挖掘与知识发现是从大规模的数据中提取潜在的、极有价值的信息的一门学科,它综合了统计学、计算机科学、管理学等前沿学科。当今社会,该技术已经得到了充分地发展和应用,并成为对海量数据进行信息提取的最为有效的方法。随着各个学科之间相互渗透的程度日益加深,数据挖掘技术与其他学科进行交叉研究已经在学术领域和商业领域得到了广泛的重视。地理事件相关研究已经在地理学得到了充分发展。大量的地理事件模型被提出,极大地丰富了该研究领域。在地理事件挖掘领域,大量专家学者针对不同的研究方向,提出了不同的地理事件挖掘算法。海洋事件挖掘是地理事件挖掘的一个重要研究方向,该研究对于研究区域海气相互作用和全球气候变化来说具有重要的科学意义和实践意义!论文分别从数据挖掘和海洋异常事件这两方面出发,在对地理事件和关联规则挖掘进行分析研究后,针对具有连续性的海洋现象,提出了两个面向海洋异常事件的关联规则挖掘算法,然后选取各种海洋环境要素异常活跃的太平洋区域为研究对象,进行实例挖掘。论文的主要工作分为如下几个部分:(1)详细介绍数据挖掘中关联规则挖掘的研究现状和地理事件的研究现状,其中地理事件的研究现状包括地理事件模型领域的研究现状和地理事件挖掘领域的研究现状。(2)阐述面向事件挖掘的时空事务表的构建流程。首先介绍在海洋环境下的与事件相关的概念与定义,然后详细介绍海洋要素异常状态提取的过程,包括提取月均距平值、海洋要素的离散化以及单一时刻海洋要素异常状态提取。最后,在以上基础上提出所构建的时空事务表,并给出对事务表进行噪音清除的方法。(3)提出处理海洋异常事件的关联规则挖掘算法。论文首先给出关联规则挖掘中经典的Apriori算法的核心思想,即根据支持度确定频繁项集,根据置信度确定强关联规则。在此基础上,结合海洋环境背景分别提出两个关联规则挖掘算法:针对共同发生事件的CE-ARMA算法和针对非同步发生事件的MAETP-ARMA算法。对于CE-ARMA算法,论文给出算法相关的概念与定义,其次,给出与Apriori算法的区别和联系,然后给出算法的详细设计流程,最后以举例的方式来说明算法的运算过程。对于MAETP-ARMA算法,论文给出算法涉及到的一些基本的概念,其次,详细地描述算法与前述算法的区别、联系,然后给出算法的设计步骤,最后,仍然以实例的方式来表述算法的运算过程。两个算法虽然具有Apriori算法的“链接-剪枝”特性,但是却解决了Apriori原型算法无法处理连续性的海洋异常事件的问题。(4)以全球气候变化和区域海洋-大气相互作用中关联关系最为密切的太平洋区域为主要研究区域,利用海表温度、海面高度异常等海洋环境要素数据进行关联规则挖掘和典型事件间的关联规则分析。最后,从时间复杂度和算法效率上对算法进行分析。
[Abstract]:Data mining and knowledge discovery is extracted from the large-scale data potential, a subject of valuable information, which combines statistics, computer science, advanced management science. Nowadays, this technology has been fully developed and applied, and has become the most effective method for massive data information extraction. With the interaction between the various disciplines of the deepening, the data mining technology to cross research with other disciplines in the academic field and business field has received extensive attention. The related research of geographical events has been fully developed in geography. A geographical event model is proposed, which greatly enrich the research field mining. In the field of geographical events, a large number of experts and scholars in different research direction, put forward different geographical event mining algorithm. Ocean mining is a geographical event One of the most important research direction of event mining, the research has important scientific significance and practical significance for the study of regional air sea interaction and global climate change! This paper respectively from the data mining and marine abnormal events of the two aspects, in the mining of geographic events and association rules analysis, according to the continuity of marine phenomena the proposed mining algorithm of two marine abnormal events association rules, and then select the various Pacific Ocean Environment Elements active as the research object, for example mining. The main work is divided into several parts as follows: (1) the research current situation and geographical events in mining association rules of data mining are introduced in detail the research status of the geographical events including research situation and research status of geographical events geographical event model in mining area. (2) describes space-time transaction table event oriented mining process. First introduced on the concept and definition of the related marine environment, and then introduces the process of marine elements abnormal state extraction, including extraction of monthly anomaly, marine elements discretization and single moment extraction. Finally the abnormal state of the ocean elements this time, the transaction table constructed on the basis of the above, and given the transaction table method for noise elimination. (3) proposed association rule mining algorithm dealing with marine abnormal events. The core idea of Apriori algorithm, this paper gives the classical association rules mining, which is based on the support degree is determined according to the frequent itemsets. Confidence determines the strong association rules. On this basis, combined with the background of the marine environment are two association rules mining algorithm: for the common occurrence of events and CE-ARMA algorithm for non synchronization The MAETP-ARMA algorithm of events. For the CE-ARMA algorithm, the related concepts and definitions, the given algorithm secondly, difference and relation are given with the Apriori algorithm, and then the detailed design process are given, finally, by way of example to illustrate the operation process of the algorithm. For MAETP-ARMA algorithm, some basic concepts related to the algorithm are given secondly, a detailed description of the algorithm and the difference algorithm, then the algorithm design steps, finally, the operation process is still in the instance of the way to express the algorithm. The two algorithm though Apriori algorithm with "link - pruning" characteristics, but are solved by the Apriori prototype algorithm can not handle abnormal events the continuity of the problem. (4) associated with global climate change and regional ocean atmosphere interaction in most Pacific region close to the study area, Association rules mining and association rules analysis between typical events are done by using marine environmental data such as sea surface temperature and sea surface height anomaly. Finally, the algorithm is analyzed from time complexity and algorithm efficiency.

【学位授予单位】：重庆交通大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP311.13;P71

【相似文献】