📊 Twitter Sentiment Analysis with Flask & Machine Learning

🚀 Overview

This project is a Flask-based web application for analyzing sentiment in tweets using various machine learning models. The application provides 11 different visualizations representing various aspects of Twitter comments, sentiment distribution, and model performance. I collected comments in English from various data sets, mainly from kaggle:

https://www.kaggle.com/datasets/jp797498e/twitter-entity-sentiment-analysis https://www.kaggle.com/datasets/kazanova/sentiment140 https://www.kaggle.com/datasets/saurabhshahane/twitter-sentiment-dataset https://www.kaggle.com/competitions/tweet-sentiment-extraction https://www.kaggle.com/datasets/cosmos98/twitter-and-reddit-sentimental-analysis-dataset https://www.kaggle.com/datasets/edqian/twitter-climate-change-sentiment-dataset

The system includes four sentiment analysis models. The models are:

• Logistic Regression
• Stacked Model (Naïve Bayes + XGBoost + Logistic Regression)
• LSTM (Long Short-Term Memory)
• ULMFiT (Universal Language Model Fine-tuning for Text Classification)

In addition, there are other sentiment classification models in the notebook and files, but they had worse accuracy, so I did not include them in the application(e.g. Random Forest, CNN,

---

🔥 Features

✅ Sentiment classification using multiple ML & Deep Learning models
✅ 11 visualizations of Twitter comments and sentiment analysis(some aspects were analysed from stop words (sw means; without stop words, other charts are with them) ) ✅ Comparison of multiple classifiers in real-time
✅ RESTful API integration for text-based sentiment analysis
✅ Web-based interface powered by Flask
✅ bilingual interface Polish and English

---

🛠 Technologies Used

• Backend: Flask, Python
• Frontend: HTML, CSS, Bootstrap, JavaScript (optional if interactive)
• Data Processing: Pandas, NumPy, Scikit-learn
• Visualization: Seaborn, matplotlib
• Machine Learning: Scikit-learn, XGBoost, Naïve Bayes, Logistic Regression
• Deep Learning: TensorFlow/Keras (LSTM), FastAI (ULMFiT)

---

📸 Application Screenshots

🏠 Main Dashboard

📊 Sentiment Analysis

---

📈 Exploratory Data Analysis (EDA) Visualizations

In addition to standard sentiment distribution plots, we performed 11 in-depth exploratory analyses to better understand the textual data. These plots illustrate how various textual features relate to sentiment:

1. Character Count vs. Sentiment
   Shows how the number of characters in each tweet correlates with different sentiment classes.

2. Word Count vs. Sentiment
   Analyzes whether the total number of words in a tweet influences its sentiment.

3. Most Frequently Used Words vs. Sentiment
   Highlights the most common words in tweets, grouped by sentiment category.

4. Most Frequently Used Words vs. Sentiment (Stopwords Removed)
   Similar to the above, but excludes common stopwords to reveal more meaningful terms.

5. Character Count vs. Sentiment (Stopwords Removed)
   Compares the character lengths of tweets after removing stopwords, across different sentiment labels.

6. Word Count vs. Sentiment (Stopwords Removed)
   Investigates how the tweet’s word count changes once stopwords are removed, segmented by sentiment.

7. Bigrams and Trigrams Analysis
   Examines the most frequent 2-word and 3-word phrases to find specific linguistic patterns tied to sentiment.

8. Emotional Profile Analysis
   Maps tweets to various emotional categories (like joy, anger, sadness) to see how they align with overall sentiment.

9. Vocabulary Diversity (Entropy)
   Measures the richness of the vocabulary within each sentiment class using an entropy-based metric.

10. Part-of-Speech Analysis
    Shows the distribution of POS tags (nouns, verbs, adjectives, etc.) across tweets with different sentiments.

11. Negation Usage Analysis
    Looks at how negation words (e.g., “not”, “no”) appear in tweets and whether they affect sentiment classification.

---

📦 Installation & Setup

Probably you will need to download Microsoft Visual C++ Build Tools and then choose MSVC v143 - Visual C++ 2019 build tools, Windows 10 SDK, C++ ATL

🔹 1. Clone the Repository

git clone https://github.com/LawyerN/Sentiment-Analysis.git
cd Sentiment-Analysis
cd src
cd app

🔹 2. Create a Virtual Environment

python -m venv venv
source venv/bin/activate   # On macOS/Linux
venv\Scripts\activate      # On Windows

🔹 3. Install Dependencies

python -m pip install --upgrade pip==25.0.1
pip install -r requirements.txt

🔹 4. Run the Flask Application

python app.py

The application will run on http://127.0.0.1:8080/. Then just try it out, choose a model then write a comment and see a result.

---

📡 API Endpoints

You can use the RESTful API to analyze sentiment directly:

🔹 1. Analyze Text Sentiment

Endpoint: /api/sentiment
Method: POST
Payload:

{
  "text": "This is an amazing project!"
}

Response:

{
  "sentiment": "Positive"
}

---

🧪 Model Performance

Model	Accuracy	Precision	Recall	F1-score
Logistic Regression	73%	74%	73%	74%
Stacked Model	74%	74%	74%	74%
LSTM	89%	88%	87%	88%

---

📊 Confusion Matrices (Model Performance Analysis)

Confusion matrices help visualize the classification performance of different models. Below are the confusion matrices for each model:

Logistic Regression Confusion Matrix

Stacked Model Confusion Matrix

LSTM Confusion Matrix

These confusion matrices show the number of correct and incorrect predictions for each sentiment category.

---

📜 Future Improvements

🔹 Add real-time Twitter scraping & analysis
🔹 Implement sentiment trend over time
🔹 Deploy the application on Heroku/AWS

---

👨‍💻 Author

Your Name – Bartosz Gdowski
GitHub: https://github.com/LawyerN
🚀 Feel free to contribute or give suggestions!

---

⭐ Like This Project?

If you found this project useful, please give it a star ⭐ on GitHub!