While many existing formal concept analysis algorithms are efficient, theyare typically unsuitable for distributed implementation. Taking the MapReduce(MR) framework as our inspiration we introduce a distributed approach forperforming formal concept mining. Our method has its novelty in that we use alight-weight MapReduce runtime called Twister which is better suited toiterative algorithms than recent distributed approaches. First, we describe thetheoretical foundations underpinning our distributed formal concept analysisapproach. Second, we provide a representative exemplar of how a classiccentralized algorithm can be implemented in a distributed fashion using ourmethodology: we modify Ganter's classic algorithm by introducing a family ofMR* algorithms, namely MRGanter and MRGanter+ where the prefix denotes thealgorithm's lineage. To evaluate the factors that impact distributed algorithmperformance, we compare our MR* algorithms with the state-of-the-art.Experiments conducted on real datasets demonstrate that MRGanter+ is efficient,scalable and an appealing algorithm for distributed problems.
展开▼
