mrhammad12/hammad-logistic-regression

1---2title: Hammad Logistic Regression3emoji: 📊4colorFrom: indigo5colorTo: green6sdk: gradio7sdk_version: 5.49.18app_file: app.py9pinned: false10license: mit11---12 13 14# 🧠 Logistic Regression from Scratch15 16**Built a Machine Learning Model from Scratch + Integrated with a Flask App**17 18This project demonstrates the implementation of a Logistic Regression model completely from scratch, with every formula and function manually implemented — no pre-built ML libraries for training or prediction! The model is integrated with a Flask web app for real-time predictions and visualization.19 20## 🚀 Demo21[![Hugging Face Spaces](https://img.shields.io/badge/🤗%20Hugging%20Face-Spaces-blue)](https://huggingface.co/spaces/your-username/logistic-regression-from-scratch)22 23Try the live demo on Hugging Face Spaces!24 25## 📋 Project Overview26 27I developed a beginner-level Machine Learning model from the ground up, implementing all mathematical formulas manually. The project includes a Flask web interface that showcases how the model performs predictions in real-time.28 29## ✅ Key Features30 31- **Custom Logistic Regression** implementation using only NumPy32- **Manual preprocessing** and feature scaling using StandardScaler33- **98% accuracy** on the Breast Cancer Wisconsin dataset34- **Flask web interface** for real-time predictions35- **Performance visualization** and prediction results36- **Complete mathematical implementation** of gradient descent and sigmoid activation37 38## 🧠 Tech Stack39 40- **Python** (core ML logic implementation)41- **Flask** (web framework)42- **NumPy, Pandas** (data manipulation and numerical operations)43- **HTML/CSS/JavaScript** (frontend interface)44- **Scikit-learn** (only for dataset loading and train-test split)45 46## 📊 Model Performance47 48- **Accuracy**: 98% on test set49- **Training Method**: Gradient Descent with L2 regularization50- **Epochs**: 200051- **Learning Rate**: 0.0152 53## 🔬 Mathematical Implementation54 55The model implements:56- Sigmoid activation function: `σ(z) = 1 / (1 + e^(-z))`57- Cost function with L2 regularization58- Gradient descent optimization59- Manual forward and backward propagation60 61## 👨‍💻 Author62 63**Hammad** - Machine Learning Enthusiast64 65## 📄 License66 67This project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.68 69---70 71*This project helped me connect theory, implementation, and presentation, reinforcing how ML models work under the hood before using frameworks like TensorFlow or PyTorch.*72 73Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference74