🦠 COVID-19 Twitter Sentiment Analysis

This project explores public sentiment surrounding the emergence of a new COVID-19 strain using Twitter data. With rising concern over rapid public reactions—such as panic buying or misinformation—the goal is to leverage Natural Language Processing (NLP) techniques to analyze real-time sentiment expressed on social media, helping health authorities make informed decisions.

---

📌 Use Case

As part of an initiative for the Ministry of Health and Family Welfare, this analysis aims to determine how people are reacting to a newly detected COVID-19 strain. Traditional survey methods are often too slow in fast-changing scenarios like a pandemic. Instead, this project taps into Twitter data to quickly assess public mood and sentiment.

---

📊 Dataset

• Source: Kaggle
• Size: 3,798 tweets and 6 features
• Nature: Each tweet is labeled with sentiment (Positive, Negative)

---

🧹 Preprocessing

Textual data was cleaned and normalized using the following NLP techniques:

• Stopword removal
• Lemmatization
• Frequency Dictionary-based token analysis

---

🤖 Model

• Algorithm: Logistic Regression
• Evaluation Metric: Accuracy on test set
• Test Accuracy: 57.89%

---

📁 Project Structure

📦 covid-twitter-sentiment
│
├── covid\_twitter\_sentiment\_analysis.ipynb  # Main analysis notebook
└── README.md                               # Project overview and documentation

---

💬 Sentiment Labels

• Positive
• Negative

(Neutral and other labels are not being used)

---

🚀 Deployment

This project currently runs as a standalone Jupyter Notebook and has not been deployed using any web framework.

---

📌 Getting Started

1. Clone the repository:

git clone https://github.com/yugmint/sentiment_analysis_twitter.git

2. Open the notebook:

jupyter notebook covid_twitter_sentiment_analysis.ipynb

---

📬 Contact

Yugendra Salunke Data Scientist 📧 [email protected] 📞 +91-8109079427 LinkedIn | GitHub Portfolio

---

⭐️ Acknowledgment

Thanks to Kaggle for the COVID-19 tweet dataset.

---

🔮 Future Enhancements

• Use more advanced models (e.g., BERT, LSTM)
• Expand to multilingual sentiment analysis
• Deploy using Streamlit or Flask for real-time, interactive analysis