Harnessing Advanced Classifiers for Sentiment Analysis on Augmented Tweets

The utilization of the internet has undergone rapid growth over the years. An increasing number of businesses find it imperative to establish an online presence to engage with a broader spectrum of potential customers. In tandem with this trend, the adoption of social media has undergone exponential expansion. Brands spanning various sectors must grasp their customers' perceptions to optimize their operations, enabling them to provide pertinent content and products to their clientele. Twitter emerges as an effective avenue for gathering insights from customers. This study curates a dataset from Twitter encompassing South African Mobile Network Operators. The acquired data is enhanced using the Valence Aware Dictionary and sEntiment Reasoner (VADER) lexicon-driven approach. The subsequent training leverages Support Vector Machine and Naïve Bayes techniques. Through the application of diverse vectorization methods, it becomes evident that both models exhibit heightened accuracy when utilizing the Term Frequency - Inverse Document Frequency (TF-IDF) technique. Among the two models, the Support Vector Machine notably achieves the highest accuracy rate, standing at 75%.