Code-lyon/ghana-sentiment-analysis

Ghana Election Sentiment Analysis Project - Complete Summary 🎯 Project Overview A comprehensive sentiment analysis system that analyzes Ghanaian Twitter discourse about the 2024 presidential candidates John Mahama and Dr. Mahamudu Bawumia using advanced Natural Language Processing (NLP) and machine learning techniques. 🏛️ Academic Context Institution: University of Energy and Natural Resources (UENR) Department: Information Technology and Decision Sciences (ITDS) Project Type: Final Year Research Project Timeline: January 2024 - November 2024 (pre-election period) 🎯 Core Objectives Main Objective To develop and evaluate an automated sentiment analysis system using NLP and machine learning to classify Ghanaian sentiments on Twitter about the presidential candidates. Specific Objectives Data Collection: Gather tweets using Twitter API focused on both candidates Text Preprocessing: Clean and normalize text data using NLP techniques Model Implementation: Train and compare multiple ML models (SVM, Random Forest, XGBoost, Logistic Regression, LightGBM, TextBlob) Performance Evaluation: Assess models using accuracy, precision, recall, F1-score Visualization: Create interactive dashboards showing sentiment trends Deployment: Build end-to-end pipeline for future political analysis 🔧 Technical Architecture Data Pipeline text Twitter API → Data Collection → Text Preprocessing → Feature Extraction → Model Training → Evaluation → Visualization Key Technologies Programming: Python 3.10 ML Libraries: Scikit-learn, XGBoost, LightGBM, NLTK, TextBlob Development: Jupyter Notebook, VS Code Visualization: Matplotlib, Seaborn, Plotly, Chart.js Deployment: Flask, Docker, GitHub Pages Machine Learning Models Support Vector Machines (SVM) - Best performer (85% accuracy) Random Forest - Strong ensemble method XGBoost & LightGBM - Advanced gradient boosting Logistic Regression - Interpretable baseline model TextBlob - Lexicon-based approach for comparison 📊 Key Findings …