IAPI QUAD-FILTER: AN INTERACTIVE AND ADAPTIVE PARTITIONED APPROACH FOR INCREMENTAL FREQUENT PATTERN MINING

Association rule mining has been proposed for market basket analysis and to predict customer purchasing/spending behaviour by analyzing the frequent itemsets in a large pool of transactions. Finding frequent itemsets from a very large and dynamic dataset is a time consuming process. Several sequential algorithms have contributed to frequent pattern generation. Most of them face problems of time and space complexities and do not support incremental mining to accommodate change in customer purchase behaviour. To reduce these complexities researchers propose partitioned and parallel approaches; but they are compromising on anyone of these. An interactive and adaptive partitioned incremental mining algorithm with four level filtering approaches for frequent pattern mining is proposed here. It prepares incremental frequent patterns, without generating local frequent itemsets in less time and space complexities and is efficiently applicable to both sequential and parallel mining.