Robust Optimization Based Extreme Learning Machine for Sentiment Analysis in Big Data

Abstract

Increasing use of social media has increased consumer interest in reading product evaluations and ratings before making a purchase. There is now a mechanism to examine natural language processing, sentiment analysis, and domain adaptation separately. A classifier trained on one set of datasets may underperform when applied to another collection of data. Therefore, it's critical to retain an open mind while experimenting with new classifiers. Reviewing datasets in big data is currently taking place. When applied to large datasets, the sentiment analysis algorithm designed for single machines or small datasets will not perform well. Robust Optimization-based Extreme Learning Machine (ROELM) is a classifier proposed in this work for sentiment analysis in massive data. ROELM is using natural wolf-like behavior to analyze an enormous review database. The single-layer hidden layer of ELM improves classification performance by one factor. This classifier's accuracy and f-measure performance have been assessed. According to the results, the suggested classifier achieves a higher level of classification accuracy than current classifiers.