Abstract: Aiming at the Fast Distributed association rules Mining (FDM) algorithm which perhaps
loses some global frequent K-item sets, an improved algorithm, named as DS-CFDM, is presented
to solve the problems of FDM. The improved algorithm introduces the ideas of dual-space and
identification bit. In order to reduce the burden of network, a center site is used to coordinate
the information transmission during the implementation of the algorithm.

Key words: distributed database, global frequent item-sets, data mining, FDM, DS-CFDM