Research And Improvement On The Algorithms Of Mining Association Rules

As the growing up of databases technique and prevalent of administrator information system, the number of unused data has been increasing at the rate of exponent. In front of large number of data, people often expect for a technology which eliminates the false numbers but retain the true values. So data mining has been proposed, which is the core of knowledge discovery in databases.Data mining is the process of extracting, identifying and discovering potentially valid, useful, previously unknown, and ultimately understandable knowledge (rules or patterns).The mining of association rules is one of the most active researches in data mining. It was proposed by Agawal in 1993. The mining of association rules is used for discovering the relation of different item-sets in databases. The algorithms of association rules fall into two categories according to the needs of producing candidate item-sets, mainly Frequent Pattern tree and Apriori_like algorithms. The main difference is that Frequent Pattern tree does not produce candidate item-sets. The latter is the algorithm needs to produce candidate item-sets.Be based on researching of data mining, we do some deep research on the mining of association rules, emphasize on Apriori algorithm, which is a classical algorithm of association rules mining. We analyze it, bring forward its shortage and propose two improvement algorithms. The main content are as follows:1) We do a summary to definition,process,technique classes and trend of Data Mining.2) We do a summary to definition,property,mining process,mining algorithms and present research of Association Rules Mining.3) We introduce the classical association rule algorithm—Apriori in detail, do some analysis on it's property, at the same time, introduce some improved algorithms based on it.4) We present Om-Apriori algorithm according to Op-Apriori, and present SMApriori algorithm according to MApriori.The main innovation points of this paper are as follows:1) We present Om-Apriori algorithm according to the properties of Op-Apriori, which improve the process of the first two frequent item-sets by MAT algorithm, improve the process of k frequent item-sets by the method in literature [34],where k≥3; Om-Apriori algorithm improve Apriori algorithm further.2) We present SMApriori algorithm according to the properties of MApriori, which make use of the property of that not all items and affairs are useful for producing frequent item-sets to shrink the matrix, SMApriori algorithm improve the MApriori algorithm on time complexity and space complexity.