Sentiment Analysis in Twitter Text by Using Different Types of Machine Learning Techniques

This research seeks to identify the best classifier using Machine Learning (ML) algorithms that can predict the polarity of a comment. The main objective of sentiment analysis is to identify the positive and negative polarities of the social forum text. To conduct this research, we collected sentiment data from the Kaggle dataset and used Natural language Processing (NLP) to classify the emotions from the Twitter text. For that, first preprocess the Twitter text by stemming and cleaning the data by removing Twitter handles, stop words, links, punctuations, numbers, and special characters. Thereafter text tokenization and normalization processes are carried out to the cleaned tweeter text. After that, the frequency word matrix has been created using a count vectorizer. Finally, the accuracy has been calculated by applying different types of classifiers to the word matrix. The accuracy obtained is 95.71% for XGB Classifier, 97.15% for the random forest classifier, 96.28 for logistic regression, 93.24% for the decision tree classifier, and 96.19% for the SVM classifier where this method gets more accuracy than the previous work.