Association Rule Mining Algorithms through Vertical and Horizontal Data Layouts: Implementation and Performance Comparison

Abstract

Data mining is used to discover interesting and previously unknown patterns from datasets.Association rule mining is a  popular and well researched method of data mining for discovering interesting relations between items in the databases. Finding frequent itemsets is an important task in data mining for extracting association rules. In this research, we have taken four association rule mining algorithms that use horizontal and vertical data formats for generation of  frequent  itemsets. we introduce a new association rule mining algorithm, intersect transaction algorithm that uses purely horizontal database layout and find the frequent itemsets by intersecting the transactions having a no. of items. We have also implemented Apriori and SplitMerge algorithms of association rule mining. Our Apriori implementation is the enhancement of the previous Apriori algorithm but uses the same property in finding the frequent itemsets as the previous Apriori algorithm uses. The Enhanced Apriori implementation is better in execution time as compared to the previous Apriori algorithm. The SplitMerge algorithms uses the split and merge technique to find the frequent itemsets. We have taken the Eclat (Equivalence CLASS transformation) algorithm that uses purely vertical data layout .I tested these algorithms on both real and synthetic datasets and then thoroughly investigate the strengths and weakness of these algorithms by carrying out several runtime experiments. It turns out that the runtime behavior of the algorithms is much more similar as to be expected