Genetic-Fuzzy Mining with MapReduce

Genetic-fuzzy mining (GFM) was proposed to train fuzzy membership functions, thus enhancing the final solution quality. However, it is quite time-consuming because of frequent database scans for fitness evaluation. In this paper, we propose a parallel algorithm with MapReduce architecture to further speed up the genetic-fuzzy mining process. In the proposed approach, the master processor randomly generates chromosomes for each item. Each chromosome is encoded in the key-value format, where the key is the item name and the value is the chromosome for corresponded item. Reduce is then utilized to execute the genetic operations. At last, experiments are conducted to show the performance of the proposed approach.