Sampling based multi-agent joint learning for association rule mining.

In order to achieve distributed data mining quickly and efficiently, this paper proposes SMAJL, a model for sampling based multi-agent joint learning which integrates sampling technology and multi-agent argumentation in the field of association rule mining. By sampling, this model can reduce the size of dataset and improve the efficiency of data mining; through joint learning from argumentation, it can effectively integrate inconsistent knowledge from different samples to improve the quality of distributed mining. We experimentally show that, in a variety of sampling strategies, SMAJL can almost achieve 90% accuracy using sample having a size of only 30% of that of original dataset.