Skip to Main Content
Many algorithms have been proposed for the discovery of association rules. The efficiency of these algorithms needs to be improved to handle real-world large datasets. Specifically, for data stored in heterogeneous and geographically distributed healthcare centers. This efficiency can be determined mainly by three factors. The way candidates are generated, the way their supports are counted and the data structure used. Most papers focus on the first and the second factors while few focus on the underlying data structures. In this paper, we present a distributed Multi-Agent based algorithm for mining association rules in distributed environments. The distributed MAS algorithm uses Bit vector data structure that was proved to have better performance in centralized environments. The algorithm is implemented in the context of Multi-Agent systems and complies with global communication standard Foundation for Intelligent Physical Agents (FIPA). The distributed Multi-Agent based algorithm with its new data structure improves implementations reported in the literature that were based on Apriori. The algorithm has better performance over Apriori-like algorithms.
Note: This article was mistakenly omitted from the original IEEE Xplore conference submission.