Social Media Sentiment Analysis

This project performs sentiment analysis on tweets using the Twitter Sentiment Analysis Dataset from Kaggle. It applies NLP techniques for data cleaning and uses models like Naive Bayes and TF-IDF with ML classifiers to classify sentiments as positive or negative.

📁 Dataset

The dataset used is: training.1600000.processed.noemoticon.csv from Kaggle's Twitter Sentiment Analysis dataset

• Sentiment: 0 = Negative, 4 = Positive (converted to 1 for binary classification)
• Text: Actual tweet text

👉 You can download it from Kaggle: Twitter Sentiment Analysis Dataset

---

🛠️ Technologies Used

• Python
• NLTK
• NumPy / Pandas
• Matplotlib
• Scikit-learn

---

🧹 Data Preprocessing

1. Drop irrelevant columns: Keep only Sentiment and Text.

2. Sentiment Mapping: 4 → 1 (positive)

3. Downsampling: Balance dataset to have equal positive and negative samples.

4. Text Cleaning:

   • Lowercasing
   • Remove stopwords & punctuation
   • Remove digits, tags, special characters
   • Lemmatization using WordNetLemmatizer

---

📊 Exploratory Data Analysis (EDA)

• Distribution of sentiments before and after balancing
• Sample visualization using histograms
• Top features contributing to sentiment classification

---

🔍 Model Building

1. NLTK Naive Bayes Classifier

• Token-based feature extraction

• Accuracy:

  • Training: ~86%
  • Testing: ~76%

2. TF-IDF + ML Classifiers

• Feature extraction using TfidfVectorizer

• Models:

  • Multinomial Naive Bayes
  • Bernoulli Naive Bayes
  • Linear SVM

• Model evaluation using accuracy and confusion matrix

---

📈 Results

Model	Accuracy (Test)
NLTK Naive Bayes	76%
Multinomial NB (TF-IDF)	~84–86%
Linear SVM (TF-IDF)	~88–90%

---

📌 Future Work

• Hyperparameter tuning using GridSearchCV
• Adding emojis/emoticons handling
• Applying deep learning models (e.g., LSTM, BERT)

---

🙌 Acknowledgements

• Kaggle Twitter Dataset by Kaggle user kazanova
• NLTK and Scikit-learn documentation

---

Would you like me to generate this as a downloadable README.md file for GitHub?