An Improved Frequent Pattern Tree Based Association Rule Mining Technique. A.B.M.Rezbaul Islam Department of Computer Engineering Ajou University Suwon, Republic of Korea damal113@yahoo.com Tae-Sun Chung Department of Computer Engineering Ajou University Suwon, Republic of Korea tschung@ajou.ac.kr Abstract—Discovery of association rules among the large number of item sets is considered as an important aspect of data mining. The ever increasing demand of finding pattern from large data enhances the association rule mining. Researchers developed a lot of algorithms and techniques for determining association rules. The main problem is the generation of candidate set. Among the existing techniques, the frequent pattern growth (FP-growth) method is the most efficient and scalable approach. It mines the frequent item set without candidate set generation. The main obstacle of FP growth is, it generates a massive number of conditional FP tree. In this research paper, we proposed a new and improved FP tree with a table and a new algorithm for mining association rules. This algorithm mines all possible frequent item set without generating the conditional FP tree. It also provides the frequency of frequent items, which is used to estimate the desired association rules. Keywords - Association rule;FP-tree;Frequent pattern data;Frequency of the items. I. INTRODUCTION Data mining is used to deal with large amounts of data which are stored in the database, to find out desired information and knowledge. Various data mining techniques such as, decision trees, association rules, and neural networks are already proposed and become the point of attention for several years. Association rule mining technique is the most effective data mining technique to discover hidden or desired pattern among the large amount of data. It is responsible to find correlation relationships among different data attributes in a large set of items in a database. Since its introduction, this method has gained a lot of attention. Association rules were first introduced in [1]. It provides information of the type of "if-then" statements. These rules are generated from the dataset and it derives from measurements of the support and confidence of each rule that can show the frequency of occurrence of a given rule. Association Analysis [1, 2, 4, 6] is the detection of hidden pattern or condition that occur frequently together in a given data. Association Rule mining techniques finds interesting associations and correlations among data set. An association rule [1,3,4,5] is a rule, which entails certain association relationships with objects or items, for example the interrelationship of the data item as whether they occur simultaneously with other data item and how often. These rules are computed from the data and, association rules are calculated with help of probability. It has a mentionable amount of practical applications, including classification, XML mining, spatial data analysis, and share market and recommendation systems. This rule measure with support to ensure every dataset treated equally in classical model. The perception of association rule mining suggests the support- confidence extent outline and condensed association rule mining to the discovery of frequent item sets. Rule support and confidence are two measures of interestingness. Association rules are regarded as appealing if a minimum support and a minimum confidence threshold is satisfied. This paper presents an efficient association rule mining technique with help of improved frequent pattern tree (FP-tree) and a mining frequent item set (MFI) algorithm. The main advantage of this mining is we can get all the frequent item set (1-item, 2-item and so on) with less affords. The outline of this paper is as follows: we discuss about some closely related work on association rule mining in Section II. Section III defines some contextual notations. At Section IV we described our key idea improved FP-tree with an example. Section V calculates the frequent items with MFI algorithm. Section VI shows the association rule mining and experimental results and Section VII concludes the paper. II. RELATED WORKS Algorithms for mining association rules from relational data have been implemented since long before. Association rule mining was first introduced at 1993 by R. Agrawal, T. Imielinski, and A. Swami [1].After that many algorithms have been proposed and developed- Apriori [7] , DHP [8] , and FP- growth [9] .The Apriori algorithm [4] uses a bottom-up breadth-first approach to find the large item set. As it was proposed to grip the relational data this algorithm cannot be applied directly to mine complex data. Another well-known algorithm is FP growth algorithm. It adopts divide-and-conquer approach. First it computes the frequent items and characterizes the frequent items in a tree called frequent-pattern tree. This tree can also utilize as a compressed database. The 978-1-4244-9224-4/11/$26.00 ©2011 IEEE