A Parallel Strategy for Generation of Fuzzy Rule Bases in Big Data Problems Using the NSGA-DO
Maykon Rocha Santana, Heloisa A. Camargo · 2019
In Fuzzy Rule Based Systems, the process of extracting rule bases from data can lead to problems related to scalability and computational cost. These situations usually occur when considering contexts where data sets have large volume, such as in Big Data problems. Parallel computing approaches with evolutionary computing techniques such as Multi-objective Genetic Algorithms can be used together to enable scalability and reduce the computational cost involved in the extraction process of Rules Bases of the Fuzzy Systems used in Big Data problems. This paper proposes a parallel modeling strategy for the generation of Fuzzy Rules Bases in Big Data problems using the NSGA-DO Multi-objective Genetic Algorithm. The aim of the proposal is to allow the increase of scalability and the reduction of computational cost when maximizing the accuracy (increase the precision) and minimizing the complexity (decrease number of rules and antecedents of the rules) of Fuzzy Systems generate in the context of Big data problems.