A comprehensive project analyzing public sentiment on Twitter related to top companies, games, and services. This was developed as part of Semester 2 - Software Tools for Artificial Intelligence and Machine Learning at Symbiosis Institute of Technology, Pune.
👥 Group Project | 📅 Academic Year: 2024–25
👨💻 Team Members: Deshna Nitin Tendulkar, Harsh Jain, Harshad Agrawal
👨🏫 Supervisor: Dr. Priyanka Deshmukh
This project explores how people react to popular brands on Twitter by performing sentiment analysis on 12,447 manually cleaned tweets. The analysis is based on four sentiment categories:
- ✅ Positive
- ⚪ Neutral
- ❌ Negative
- 🚫 Irrelevant
Unlike many projects that use black-box sentiment models like VADER, this study focuses entirely on statistical interpretation and data visualization, making the results more transparent and explainable.
- Analyze real-world Twitter sentiment using manual data and statistical tools.
- Apply descriptive statistics, correlation, regression, and hypothesis testing.
- Use Python (Pandas, Seaborn, Matplotlib) and Excel to produce clean, clear visualizations.
- Avoid reliance on automated NLP tools to ensure human-centered interpretability.
- ✅ Data preprocessing and validation in Python
- 📈 Created statistical visualizations (bar plots, heatmaps, scatter plots)
- 🔬 Conducted regression, correlation, and t-tests
- 📄 Contributed to report writing and presentation design
- 💡 Assisted in final polishing and presentation
- 🐍 Python (Jupyter Notebook)
- 📊 pandas, numpy, matplotlib, seaborn, scipy
- 📁 Microsoft Excel
- 📝 Manual sentiment labeling
- 📄 PDF and PPT documentation
- Sentiment Distribution:
- Negative: 3,757 (42%)
- Positive: 3,472 (33%)
- Neutral: 3,053 (25%)
- Brands like Cyberpunk 2077 and Assassin’s Creed received high positive feedback.
- Call of Duty, FIFA, and World of Warcraft had high negative sentiment.
- Correlation between Negative and Positive: –0.55
- Linear Regression R²: 0.08
- T-Test p-value: 0.347 (no significant difference between means)
- 📊 Stacked bar charts per entity
- 🔥 Heatmaps of sentiment correlations
- 📈 Regression plots for sentiment relationships
- 🧩 Pair plots across all sentiment categories
- 🕸️ Radar plots showing entity-specific sentiment balance
- 📣 Brand reputation analysis
- 🎯 Marketing feedback monitoring
- 📊 Public opinion research
- 🚨 Crisis management insights
- 🧠 Social media analytics dashboards
This project is for educational and academic purposes. Feel free to use or reference it with credit to the team.
Harshad Agrawal
B.Tech in Artificial Intelligence and Machine Learning
Symbiosis Institute of Technology, Pune
🔗 GitHub