Learning to Detect Human Emotions in Digital World by Integrating Ensemble Voting Classifiers

Abstract

Due to the expansion of world of the internet and the quick acceptance of platforms for social media, information is now able to exchange in ways never previously imagined in history of mankind. A social networking site like Twitter offers a forum where people may interact, discuss, as well as respond to specific issues via short entries, like tweets of 140 characters and fewer. Users may engage by utilizing the comment, like and share tabs on texts, videos, images and other content. Although platforms for social media are now so extensively utilized, individuals are creating as well as sharing so much information than shared before, which can be incorrect or unconnected to reality. It is difficult to identify erroneous or inaccurate statements in textual content autonomously and find emotions of people. In this paper, we suggest an Ensemble method for sentiment and emotion analysis. Different textual features of actual and Emotion and sentiment have been utilized. We used a publicly accessible dataset of twitter sentiment analysis that included total 48,247 authenticated tweets out of 23,947 of which were authentic positive texts labeled as binary 0s and 24,300 of which were negative texts labeled as binary 1s. In order to assess our approach, we used well-known (ML) machine learning techniques, these are Logistic Regression (LR), AdaBoost, Decision Tree (DT), SGD, XG-Boost as well as Naive Bayes. In order to get more accurate findings, we created a multi-model sentiment and emotion analyzing system utilizing the ensemble approach and the classifiers stated above. Our recommended ensemble learner method outperforms individual learners, according to an experimental study.