PARALLELIZATION OF BAT ALGORITHM USING HADOOP MAPREDUCE

Abstract

Bat algorithm is among the most popular meta-heuristic algorithms for optimization. Traditional bat algorithm work on sequential approach which is not scalable for optimization problems dealing with BIG DATA such a scrawled documents, web request logs, commercial transaction information therefore parallelizing meta-heuristics to run on tens, hundreds or even thousands of machine to reduce runtime is required. In this paper, we introduced two parallel models of BAT algorithm using MapReduce programming model proposed by Google. We used these two models for solving the Software development effort optimization problem. The experiment is conducted using Apache Hadoop implementation of MapReduce on a cluster of 6 machines. These models can be used to solve problems with large search space, dimension and huge computation by simply adding more hardware resources to the cluster and without changing the proposed model code.