A NEW PARALLEL FRBCS MODEL BASED ON WANG-MENDEL AND PARTICLE SWARM OPTIMIZATION ALGORITHMS

Fuzzy rule-based classification systems (FRBCS) have emerged as effective tools for classification and pattern recognition, requiring data mining algorithms to create a fuzzy rule base (FRB) in a more accurate way. This study proposed a new parallel FRBCS model based on Wang-Mendel (WM) and particle swarm optimization (PSO) algorithms called the WPF-BigData model, which was designed using the MapReduce framework. The proposed model generated and optimized the FRBS from the perspective of accuracy and reduced the elapsed time of big data classification, while serial approaches could not cope with large datasets in the era of big data. The experimental results demonstrated that this model performed competitively compared with other FRBCs models. Furthermore, it measured the speedup of WPF-BigData model with a dataset containing up to 5.4 million instances. The results confirmed that the parallel model was able to solve the big data classification problem with higher classification accuracy compared with the parallel WM algorithm and Chi-BigData method in an acceptable amount of time.