Improved Data Clustering using Moth Flame Optimization in Big Data

Abstract

The popularity and attraction towards internet technology, social media, etc. had resulted in generation of huge amount of data. To address data mining and information retrieval, clustering had emerged as an efficient way to partition different categories of data. In this respect, authors have improved and integrated cosine similarity at the initial stage to cluster similar documents. The clusters so obtained, are optimized using Moth Flame Optimization (MFO) and optimized clustered are fed to Deep Neural Network (DNN). The author’s uses validation propagated architecture to ensure the least significance of outliers in the work. DNN generated the classified results from the similarity based clustering of the data. The experimentation performed using large dataset retrieved from US Census Demographic data resulted in better precision, recall and f-measure. The comparative analysis further proved that the clustering of data achieved using proposed data clustering is 3.27% higher than the conventional techniques