An Improved FP-Tree Algorithm with Relationship Technique for Refined Result of Association Rule Mining

Construction and development of classifier that works with more accuracy and performs efficiently for large database is one of the key tasks of data mining techniques. Secondly training dataset repeatedly produces massive amount of rules. It’s very tough to store, retrieve, prune, and sort a huge number of rules proficiently before applying to a classifier. In such situation FP is the best choice but problem with this approach is that it generates redundant FP Tree. A Frequent pattern tree (FP-tree) is type of prefix tree that allows the detection of recurrent (frequent) item set exclusive of the candidate item set generation. It is anticipated to recuperate the flaw of existing mining methods. FP – Trees pursues the divide and conquers tactic. In this thesis we have adapt the same idea for identifying frequent item set with large database. For this we have integrated a positive and negative rule mining concept with frequent pattern algorithm and correlation approach is used to refine the association rule and give a relevant association rules for our goal. Our method performs well and produces unique rules without ambiguity.