Twitter Sentiment Analysis Project

A modular machine learning pipeline for classifying tweet sentiments as positive or negative, using robust text preprocessing, feature engineering, and model comparison. This project demonstrates practical NLP techniques and evaluates multiple classifiers to identify the most effective model for real-world sentiment prediction.

---

🧠 Objectives

• Clean and preprocess raw Twitter data for sentiment classification.
• Extract meaningful features using Bag-of-Words (BoW) and TF-IDF.
• Train and evaluate multiple ML models to identify the best performer.
• Visualize insights from the data (hashtags, word clouds).
• Generate predictions on unseen test data using the top-performing model.

---

⚙️ Workflow Overview

graph TD
A[Load Data] --> B[Preprocess Tweets]
B --> C[Feature Extraction: CBoW & TF-IDF]
C --> D[Train Models: LogReg, XGBoost, Decision Tree]
D --> E[Evaluate via F1-Score]
E --> F[Visualize Results]
F --> G[Predict on Test Data]

Loading

---

📦 Requirements

Install dependencies using:

pip install -r requirements.txt

Also download NLTK tokenizer:

import nltk
nltk.download('punkt')

requirements.txt includes:

• pandas
• numpy
• nltk
• scikit-learn
• xgboost
• matplotlib
• seaborn
• wordcloud

---

🚀 How to Run

1. Clone this repository:

   git clone https://github.com/SatChittAnand/Twitter-Sentiment-Analysis.git
   cd Twitter-Sentiment-Analysis

2. Place the datasets train_SentimentAnalysis.csv and test_SentimentAnalysis.csv in the root directory.

3. Run the script:

   python sentimentanalysistwitter.py

The script will:

• Preprocess and vectorize the data
• Train and evaluate models
• Generate visualizations
• Save predictions to predictions.csv

---

🤖 Models Implemented

Model	Feature Technique	Evaluation Metric
Logistic Regression	BoW, TF-IDF	F1-Score
XGBoost Classifier	BoW, TF-IDF	F1-Score
Decision Tree	BoW, TF-IDF	F1-Score

---

📊 Results

• Logistic Regression with TF-IDF achieved the highest F1-score.
• Visual comparisons via point plots highlight model-feature performance trade-offs.
• Word clouds and hashtag frequency plots offer intuitive insights into tweet content.

---

📁 Project Structure

twitter-sentiment-analysis/
│
├── sentimentanalysistwitter.py       # Main script
├── train_SentimentAnalysis.csv       # Training dataset
├── test_SentimentAnalysis.csv        # Test dataset
├── predictions.csv                   # Output predictions
├── requirements.txt                  # Dependencies
└── README.md                         # Project documentation

---

🌟 Highlights

• ✅ Modular design for easy extension
• 📈 Visual insights into tweet content and trends
• 🔁 Reproducible and scalable for larger datasets
• 🧩 Easy to integrate into real-time sentiment monitoring systems

---

🤝 Contributing

Pull requests are welcome! For major changes, please open an issue first to discuss what you’d like to modify.

---

📜 License

This project is licensed under the MIT License.

---

🙌 Acknowledgments

• NLTK
• Scikit-learn
• XGBoost
• Seaborn
• WordCloud

---