Hybrid optimization-based deep learning classifier for sentiment classification using review data

The study of opinions conveyed in a text format is examined using sentiment analysis. Analyzing the sentiment plays a fundamental part in the sentiment orientation of reviews and extracting necessary information. Numerous existing sentiment classification approaches are designed with lesser classification performance. Hence, it is vital to design a sentiment classification model that poses the capability for classifying the opinions expressed in review data. An effective sentiment classification method, named exponential poor–rich optimization-based gated recurrent unit, is introduced in this research for classifying the sentiment using reviews. The developed exponential poor–rich optimization algorithm is newly derived by the merging of poor–rich optimization and exponential weighted moving average and poor–rich optimization. With a deep residual network, necessary features are extracted, and then, feature fusion is done by RV coefficient. The process of classification of sentiment is completed using a gated recurrent unit classifier. The training of the gated recurrent unit classifier is done by the exponential poor–rich optimization algorithm. The devised technique outperformed various existing sentiment classification approaches and attained better performance by achieving a higher true positive rate of 0.934, true negative rate of 0.901, and accuracy of 0.912.