Applying NLP and Machine Learning for Sentiment Classification Across Multiple Twitter Datasets

Abstract

Social networks have appeared as the main opinion- sharing and discussion-enabling resources in the last ten years. At the same time, the development of machine learning (ML) and natural language processing (NLP) technologies has allowed new approaches to analyzing the huge quantities of data users create. This research uses data loading, class imbalance handling, text preprocessing and tokenization, sentiment analysis, and model assessment techniques to analyze the sentiment of the tweets across two datasets. Using metrics like accuracy, precision, recall, and F1 score the study reveals that SVM and Logistic Regression are the most suitable machine- learning models for this purpose. SVM attained an accuracy of 90% for training and 77% for testing. In comparison, Logistic Regression showed 83% for training and 78% for testing for Dataset 1, and for Dataset 2 SVM attained 98% training and 86% for testing. In comparison, Logistic Regression showed balance between training and testing achieving 93% in training and 87% in testing.