Comparative Evaluating of Machine Learning Classifiers through Sentiment Analysis to Detect Emotions Using TF-IDF

Social networks are experiencing a significant surge in the demand for text mining. A growing number of persons are participating in online text analysis. As Facebook and Twitter gain popularity, a growing number of users are composing extensive notes on these platforms. The analysis of these comments holds paramount importance for various business applications. Sentiment Analysis (SA), a technique within Natural Language Processing (NLP), plays a pivotal role in discerning the emotions conveyed in reviews and sentiments. This paper introduces the development of machine learning model using TF-IDFVectorizer (Term Frequency Inverse Document Vectorizer) Frequency for feature extraction technique. The aim is to predict the emotional activities of user comments through sentiment analysis. The primary objectives of this proposal are threefold. Initially, we collected a dataset and categorized the dataset for positive and negative. Secondly, we conducted a comprehensive comparison of seven classifiers such asDecision Tree, K Nearest Neighbor, Naïve Bayes, AdaBoost, XGBoost, Multi layer Perceptron, and Proposed Random Forest Model. Finally, we present the efficacy of our proposed model of Random Forest, showcasing state-of-the-art result in Borderland Emotion Dataset.