Incremental Association Rule Mining Based on Matrix Compression for Edge Computing

A growing amount of data is being generated, communicated and processed at the edge nodes of cloud systems; this has the potential to improve response times and thus reduce communication bandwidth. We found that traditional static association rule mining cannot solve certain real-world problems with dynamically changing data. Incremental association rule mining algorithms have been studied. This paper combines the fast update pruning (FUP) algorithm with a compressed Boolean matrix and proposes a new incremental association rule mining algorithm, named the FUP algorithm based on a compression matrix (FBCM). This algorithm requires only a single scan of both the database and incremental databases, establishes two compressible Boolean matrices, and applies association rule mining to those matrices. The FBCM algorithm effectively improves the computational efficiency of incremental association rule mining and hence is suitable for knowledge discovery in the edge nodes of cloud systems.