ADMiner: An Incremental Data Mining Approach Using a Compressed FP-tree

In real world applications, most transaction databases are often large and constantly updated. Current data mining algorithms face the problem of processing a large number of transactions in dynamic environments. Since memory space is limited, it is critical to be able to use available storage efficiently and to process more transactions. In this paper, we propose an improved data structure of a compressed FP-tree to mine frequent itemsets with greater efficiency. Use of our method can minimize the I/O overhead, and, more importantly, it can also perform incremental mining without rescanning the original database. Our experimental results show that the method we propose not only requires less memory, but also performs incremental mining more efficiently.