Publication

Abstract : In present days, since the size of the datasets is being increased day by day, extracting knowledge from this huge data set has become a very big issue. Particularly speaking about transactional data set, there were many approaches to find the frequent item sets. Many classical algorithms and techniques of Apriori have been proposed and used to find the frequent item sets from the huge transactional dataset. The classical algorithms scans the database repeatedly for finding frequent item sets which generally takes more execution time. In this paper we propose a new improved matrix approach with subset count to find the frequent item sets. First the given data is converted to matrix form where the items with less than minimum support count and duplicate transactions are removed, secondly each transactions are scanned for item sets and if this item set is not present in frequent item sets list the we find the subsets of this item set and add them to subset count list by incrementing the count of a particular subset. If the count of any subset is greater than minimum support then the subset is added to frequent item sets list. In this paper we have compared the existing apriori algorithm with improved matrix approach based on execution time.