Moth Flame Optimization and Deep Learning Based Digital Review Sentiment Mining

Abstract

Organization providing service or product depends on its customer reviews. Collection of such review through online platform provides digital raw data. Extracting information about product review depends on sentiment analysis from the user text review. Language processing is an complex task in case of review analysis as user words and sentiment relation is different for various context. This work has proposed a sentiment analysis hybrid model that combines genetic algorithm and deep neural network techniques.  Genetic algorithm reduces the dimension of input data by clustering terms into two class first is positive sentiment oriented class and other is negative sentiment oriented class. Clustering is done by moth flame optimization algorithm. Learning model linear regression take input of sentiment orient class terms and apply deep feature extraction operators convolution, max-pooling for getting more effective results. Experimental work was done on real dataset of amazon reviews with different testing size. Evaluation parameter values show that proposed model has increased accuracy of work by filtering data.