Fuzzy rule based multi class sentiment analysis using hybrid nature inspired machine learning technique

Due to the increased availability of the Internet and various social media, the sharing of views or comments about any issue has increased. These reviews act as an important resource for understanding the feelings of customers and concerns related to the product and thus, need to be processed properly. Sentiment analysis plays a vital role in processing the reviews and obtaining information that can be used further. Many authors prefer to perform the binary sentiment task by classifying the reviews into either negative or positive classes. The present paper chooses multi-class classification, where the reviews are classified into five different classes. To perform the binary classification task, two sets of hybrid nature-inspired techniques are used i.e., firstly Particle Swarm Optimization (PSO) and Artificial Neural Network (ANN); and secondly, Flower Pollination Algorithm (FPA) and Artificial Neural Network (ANN). PSO and FPA are used for feature selection and ANN then processes the selected features for binary classification. The reviews that are classified correctly go through a Fuzzy rule-based system for multi-class classification. For classification purposes, two movie review datasets namely, IMDb and Polarity dataset are considered.