Abstract: Present study proposes an algorithm for finding frequent itemsets. Algorithm uses a novel approach to the insignificant transactions dynamically. It divides the tuples of the database to be mined intelligently in clusters. During a particular pass only those clusters that seem to be statistically useful are to be scanned and as a consequence all insignificant tuples will be filtered out dynamically. Further, the algorithm is based on a vertical data layout and offers flexibility during mining process. Experiments have been performed on real databases and the results have been presented. The results show that by removing false frequent items and insignificant transactions dynamically, the performance of association rule-mining algorithms can be improved. It has also been observed that the performance gap increases with the large size of database and/or when there exist prolific size frequent itemset in the database at the given value of minimum support.