Sentiment Polarity Classification Using Conjure of Genetic Algorithm and Differential Evolution Methods for Optimized Feature Selection

Abstract

Background: Sentiment Analysis (SA) has a big role in Big data applications regarding consumer attitude detection, brand/product positioning, customer relationship management and market research. SA is a natural language processing method to track the public mood on a specific product. SA builds a system to collect/examine opinions on a product in comments, blog posts, re- views or tweets. Machine learning applicable to Sentiment Analysis belongs to supervised classifi- cation in general. Methods: Two sets of documents, training and test set are required in machine learning based classification: Training set is used by classifiers to learn documents differentiating character- istics; it is thus called supervised learning. Results: Test sets validate the classifier’s performance. Se- mantic orientation approach to SA is unsupervised learning because it requires no prior training for mining data. It measures how far a word is either positive or negative. This paper uses a hybrid GA- DE optimization technique for sentiment classification to classify features from movie reviews and medical data. Conclusion: Our research has enhanced the variables on learning rate as well as momentum values which are optimized by genetic approach that in turn improve the accuracy of classification procedure.