Sentiment Analysis based on Feature Selection Classification model
D. Elangovan, V Subedha · Research Square · 2022
Abstract As the use of social media and other online forums for exchanging ideas and exchanging opinions has grown, so has the field of sentiment analysis. Many people use social media like Twitter and Facebook to share their thoughts and ideas with the rest of the world. Opinion mining or sentiment analysis focuses on classifying and predicting a target's opinion. Text documents or sentences can be categorized based on whether they represent a favorable or negative view on a certain subject. Text categorization may seem like an easy process compared to sentiment analysis, but numerous difficulties have prompted many studies into this area. The use of Firefly with Levy and Multilayer Perceptron approaches combined with a feature selection-based classification model to automate sentiment analysis has been proposed. Online product reviews can be improved by incorporating feature selection and classification into the SA model presented in this research. Feature extraction from online product evaluations was accomplished using the firefly (FF) algorithm, while sentiment classification was accomplished using a multi-layer perceptron (MLP). Four datasets are used in the experiment, and the outcomes are evaluated according to various criteria. As a result of these tests, it can be concluded that the FF-MLP model provided here has the best classification performance over the whole dataset tested.