The Research And Application On The Algorithms Of Mining Association Rules

With the extensive use of database, the scale of it expands quickly. It is necessaryto analyze these data and explore valuable information from them. Data mining is atechnique that aims at analyzing data and comprehending and discovering the hiddenknowledge of data. Data mining has already aroused the information industrial fieldenormous interest and will have a wide application prospect in the future. Being anextremely essential research topic in data mining, association rules mining is widelyapplied in various fields. Association rules may both examine the knowledge patternformed for a long time in the profession and discover the secret new rules. Thediscovery, comprehension and application of association rules are important means ofaccomplishing the task of data mining. Therefore, the research of association rulesmining is of great importance in both theoretical realm and realistic realm.This paper conducts a systematic research on the association rules mining mainlyincluding the content as follows:1. The analysis and study of data mining technology. This paper gives a briefretrospect on the history of data mining technology. Based on the fundamentalconcepts of data mining technology, a detailed classification and summary on theobjects of data mining, detectable patterns and frequently used techniques have beencarried out. In succession,this paper analyses and studies the current status of datamining technique and discusses its hot research fields.All of the above provide a basefor the overall expanding of this paper.2. The study and analysis of association rules. A certain extent introduction onassociation rules mining and its algorithms as well as its species are given on theground of current research of association rules documents.And the fundamentalconcepts and essential properties of association rules are explained in detail. Inaddition, the typical mining algorithm and its fundamental ideas of association rulesare analyzed and studied in this paper. All kinds of optimized techniques designed topromote the algorithm's efficiency are also studied and discussed in detail here and atthe same time their merits and defects are analysed objectively.3. The evaluation criteria of association rules are based on support andconfidence in existing association rule mining algorithms.However,the minedassociation rules with high support and confidence are useless sometimes.This paperinvestigates the limitation of the present criterion of the evaluation of association ruleand introduces another measurement standard of association rule—RelativeSupport( S R), making Support, Confidence and Relative Support together as theevaluation criteria of valid association rules.4. The classical algorithm of association rules mining——Apriori is analyzed inthis paper. An improved and faster mining algorithm is put forward—MM_Apriori(Multiply_Matrix_Apriori).In this algorithm,we can get the frequent 2-itemsets bymultiplying two matrixes. Both the theory and tests prove that this improvedalgorithm has favorable performance.5. A study of association rule mining used in commerce is conducted. A saledatabase of a supermarket in life has been chosen as a mining object. By usingMicrosoft VC++ on Microsoft Windows 2000 Server,Microsoft SQL Server 2000 andthe MM_Apriori algorithm, the potential relations between the commodities areworked out, making theoretical research moving up to the practical application.