An efficient approach for sentiment analysis in a big data environment

Abstract

Sentiment analysis is a very substantial area of research. Numerous studies have examined the subject in recent years. It has rapidly gained interest by reason of the unusual volume of opinionated data on the Internet. Significant research has been accomplished to study sentiment by resorting to diverse machine learning techniques. Nevertheless, the downward trend of the accuracy rates in these studies often impacts the approach’s efficiency. With the aim of surmounting this obstacle, we introduce an efficient technique for sentiment mining in big data context. The data collected are cleaned using a preprocessing data mining technique before proceeding to the selection of the optimal features with the use of a versatile approach of greedy algorithms, called Carousel greedy, combined with a bio-inspired metaheuristic algorithm. The classification is subsequently performed by Cat Swarm Optimization Based Functional Link Artificial Neural Networks classifier and the performance of the approach is discussed through experimental results.