Study Of Association Rules Algorithms In Data Mining

With the development of computer and information technology, data mining hasbecome the focus of attention, widely used in all walks of life. Association rule miningis one of the most important parts in data mining. It obtain the relationship betweenfrequent itemsets by find all frequent itemsets in transaction database and dig outvaluable knowledge and information for users. Association Rules Mining plays animportant role in insurance coverage design, stock market analysis, network dataprocessing and other fields. With the development of database applications, the speedof data collection and storage increases ceaselessly, the traditional association rulemining algorithm has been unable to adapt to such changes, so the study of a highperformance data mining algorithm is very necessary.This thesis analyzed Apriori and FP-Growth in association rule algorithm andresearched them in detail, on this basis of the algorithm come up with two improvedalgorithms. The main work is as follows:Study of the basic concept of data mining technique, the process of data miningand association rules mining concept, basic principles and main research direction indetail.Analysis and research on data mining association rules of the classic Apriorialgorithm, The algorithm has short-coming in the implementation process of scanningdatabase too many numbers and produce a large number of candidate itemsets defects,we put forward a kind of algorithm based on the improved algorithm, the improvedalgorithm in mining process can removed Non frequent itemsets timely and onlyneeds to scan the database once, the follow-up work is coped in memory, so in timeand space efficiency has greatly improved.In-depth study of the frequent pattern growth algorithm FP-Growth, the algorithmneed not to produce candidates and widely used in the current mining frequent itemsetsalgorithm, but FP-Growth algorithm can not effectively mining large databases, andthe time and space complexity is higher. Aiming at the deficiency of the algorithm, thisthesis improved the original algorithm. The improved algorithm improve miningefficiency by adopted cancel redundant item, the decomposition database method formining association rules, at the same time greatly meet the needs of large databasemining.