New Distributed-TOPSIS Approach for Multi-Criteria Decision-Making Problems in a Big Data Context

Abstract

Abstract Nowadays, the online environment is extra information-rich and allows companies to offer and receive more and more options and opportunities in multiple areas. Thus, decision-makers have abundantly available alternatives to choose from the best one or rank from the most to the least preferred. However, in the multicriteria decision-making field, most tools support a limited number of alternatives with as narrow criteria as possible. Decision-makers are forced to apply a screening or filtering method to reduce the size of the problem, which will slow down the process and eliminate some potential alternatives from the rest of the decision-making process. Implementing MCDM methods in high-performance parallel and distributed computing environments becomes crucial to ensure the scalability of multicriteria decision-making solutions in Big Data contexts, where one can consider a vast number of alternatives, each being described on the basis of a number of criteria. In this context, we consider TOPSIS one of the most widely used MCDM methods. We present a parallel implementation of TOPSIS based on the MapReduce paradigm. This solution will reduce the response time of the decision-making process and facilitate the analysis of the robustness and sensitivity of the method in a high-dimension problem at a reasonable response time. Three multicriteria analysis problems were evaluated to show the proposed approach's computational efficiency and performance. All experiments are carried out within GCP's Dataproc, a service allowing the execution of Apache Hadoop and Spark tasks in Google Cloud. The results of the tests obtained are very significant and promising.