A hybrid neuro-fuzzy technique to overcome clustering approach issues in big data

Abstract

MapReduce is regarded as one of the major enabling methods in the big data community for handling the ever-increasing demands on computer resources imposed by enormous datasets. Currently, Google's MapReduce or its open-source equivalent Hadoop are well-known solutions for these types of applications. Several techniques are developed based on Hadoop such as apriori(M), HFUPM, apriori, EFUPM. The conventional algorithms of clustering schemes suffer from many challenging issues as clustering error, and computational complexity. To overcome the challenges of existing techniques we develop a novel approach to data pre-processing to extract the significant information. After pre-processing the data is processed applying neuro fuzzy-based scheduler to obtain semantic relationship between user queries. The fuzzy logic considers query time, query length, query expiry, total queries, CPU usages, and task activity. The proposed approach is implemented using a Hadoop platform. The comparative study shows a significant improvement in MapReduce performance for huge dataset.