Two Phases Association Rule mining of Remote Sensing Images based Partition and BitSet

This paper presents TP-PB algorithm. It applies two phases association rule mining one based partition and Bitset for massive remote sensing images data. Firstly, this algorithm divides massive database into several independent blocks logically. Boolean values are stored in the compressed BitSet in each of block. It generates frequent itemsets by Bitset logical AND operation replaces database scans for each of block. Then it unifies the frequent itemsets and generates global candidate itemsets for the whole data. At last it calculates the min-support of these global candidate itemsets and generates global frequent itemsets for the whole data. By dividing the massive data into a number of blocks, it enhance generic application of algorithm ; Besides it increases algorithm efficiency that database scans is replaced by Bitset logical AND operation in each of block, especially for massive remote sensing image data. And the algorithm has been applied to remote sensing images mining association rules.