<b>Twitter Sentiment Analysis of Nigeria's 2025 Tax Reform: A Comparative Study of Pre-trained Transformer Models</b>

This study presents a comprehensive sentiment analysis of Nigerian Twitter discourse surrounding the 2025 Tax Reform Act, employing a comparative machine-learning framework to quantify public perception of a landmark fiscal policy. A dataset of 2,299 manually annotated tweets was scraped between July 1 and December 31, 2025; data collection concluded at the close of this window, with all modelling, analysis, and manuscript preparation completed in early 2026. Three classification models were developed and evaluated: Logistic Regression with TF-IDF features, a Bidirectional Long Short-Term Memory (BiLSTM) network, and a fine-tuned cardiffnlp/twitter-roberta-base-sentiment-latest transformer. Model performance was evaluated using Macro-F1 as the primary metric to address severe class imbalance (87.9% Negative). The fine-tuned RoBERTa model achieved the highest Macro-F1 of 0.6073, significantly outperforming the BiLSTM (0.4763, p=0.0000) and demonstrating a large practical effect over Logistic Regression (0.4376, Cohen's d). Full-corpus sentiment analysis revealed a decisively critical public reception: approximately 87.94% negative, 8.69% positive, and 3.37% neutral (95% Wilson CI confirmed). Provision-level analysis using keyword tagging and chi-square testing (χ²=67.23, p<0.001) established that Governance & Accountability (98.2% negative) and Economic Burden provisions (94.1% negative) generated the strongest opposition, while VAT & Consumption Tax attracted the most mixed reactions (19.6% positive). These findings constitute the first data-driven characterisation of Nigerian public sentiment toward the 2025 Tax Reform Act and provide actionable intelligence for policymakers and communication strategists. My Supervisor at DataraFlow, Winner Emeto, is a dynamic and cross-functional technologist with a strong foundation in software engineering, artificial intelligence, and data science. DataraFlow is Africa's accountability-driven talent accelerator. They don't just teach Data Science, Machine Learning, and Generative AI; they build proof that graduates from the program show up every day, for six months.