P-tree oriented association rule mining of multiple data sources

摘要

A prominent research area in data mining field is (ARM). As distributed databases emerged, need to mine different patterns across them become necessary and hence distributed ARM algorithms were desired. But these algorithms increased communication complexity and overhead. This paper proposes a new algorithm P-tree oriented distributed association rule mining (PDAM) for mining association rules from distributed databases. This algorithm enables a quicker computation of support counts of item sets. P-tree, a special kind of data structure is used in the algorithm which holds the transactional data. These tree data structures do effective storage of data by employing lossless compression techniques. Message exchange optimisation is proposed in this paper. Both the database scans as well as message exchanges are reduced by the proposed method. It would also reduce the size of average transactions, data sets and message exchanges.