🌪️ Disaster Tweet Sentiment Classification

This project uses Machine Learning and Natural Language Processing (NLP) to classify tweets as disaster-related or non-disaster-related.
It was developed as part of the APSCHE Major Project in Data Science.

📘 Overview
During disasters, people often post information and alerts on social media.
The goal of this project is to automatically detect whether a tweet truly indicates a disaster event.
By applying machine learning models to real-world Twitter data, this system helps authorities and organizations respond more efficiently during emergencies.

📊 Dataset
The dataset consists of tweets labeled as “Disaster” or “Not Disaster”.
Each tweet includes:

• Tweet text
• Keywords
• Location (when available)
• Label (1 → Disaster, 0 → Not Disaster)

Data preprocessing included:

• Removing URLs, mentions, hashtags, and punctuation
• Tokenization and stop-word removal
• Lemmatization
• Converting text into numerical features using TF-IDF Vectorization

⚙️ Methodology

1. Data Collection – Imported the tweet dataset from CSV format.
2. Data Cleaning – Removed noise and handled missing values.
3. Text Preprocessing – Tokenized, removed stopwords, and lemmatized words.
4. Feature Extraction – Converted processed text into TF-IDF vectors.
5. Model Training – Trained multiple ML models.
6. Evaluation – Compared model accuracy and confusion matrices.

🧠 Models Used

• Naive Bayes Classifier
• Logistic Regression
• K-Nearest Neighbors (KNN)

Each model was evaluated using metrics such as:

• Accuracy
• Precision
• Recall
• F1-Score

📈 Results

Model	Accuracy	Precision	Recall	F1-Score
Naive Bayes	~86%	0.85	0.84	0.84
Logistic Regression	~88%	0.87	0.86	0.86
KNN	~80%	0.78	0.77	0.78

Logistic Regression gave the best overall accuracy and balance between precision and recall.

💡 Future Scope

• Implement deep learning models like LSTM or BERT for improved accuracy.
• Create a real-time tweet monitoring dashboard for live disaster classification.
• Expand the dataset to include regional languages and multilingual tweets.

🧰 Tools & Technologies

• Language: Python
• Libraries: Scikit-learn, Pandas, NumPy, Matplotlib, NLTK
• Environment: Jupyter Notebook

👩‍💻 Author
Pendyala Geetha Sri B.Tech – Computer Science and Engineering (Data Science)
APSCHE Major Project

---