Mining rules in large databases with a threads-based GA

Rule mining allows to discover interesting rules in large databases. The process is time-consuming and I/O intensive. Several approaches are proposed to deal with that problem. however, concurrency is not often considered enough for high-performance data mining. In this paper, we propose a threads-based genetic algorithm for rule discovery, namely MT - RMGA. We evaluated it on the Nursery School public domain data set available from. the UCI Repository of Machine Learning databases. The results demonstrate that tire algorithm allows to discover high quality rules efficiently. Furthermore, a multi-threaded implementation allows to enhance the performance by a factor of over 14.5%.