🧠 Machine Learning Systems Portfolio

Production-Ready ML Engineering Projects by Hubert Domagała

🌟 Try the Live Demo → - Interactive Cancer Detection System

🎯 Portfolio Overview

This repository showcases end-to-end machine learning system development — from exploratory data analysis to production-ready deployments. Each project demonstrates software engineering best practices, scalable architectures, and real-world problem-solving.

🏆 Featured Projects

Project	Domain	ML Techniques	Status	Highlights
🏥 Cancer Detection	Healthcare	Classification, Ensemble	✅ Production	96.7% accuracy, FastAPI, Zero false negatives
💳 Fraud Detection	Finance	Anomaly Detection, Feature Engineering	📊 Analysis	SMOTE, Cost-sensitive learning
✍️ Digit Recognition	Computer Vision	PCA, Neural Networks	📊 Analysis	Multi-model comparison
🍯 Honey Production	Agriculture	Time Series Regression	📊 Analysis	Trend analysis, Forecasting
🏴 Flag Analysis	Data Mining	Multi-class Classification	📊 Analysis	UCI dataset, EDA
🍇 Raisin Classification	Agriculture	Clustering, Classification	📊 Analysis	Feature analysis
💰 Income Classification	Economics	Binary Classification	📊 Analysis	Socioeconomic analysis
🏥 Medical Insurance	Healthcare	OOP Design, Regression	📊 Analysis	Clean code architecture

Legend: ✅ Production (API + Tests) | 📊 Analysis (Notebooks)

🚀 Quick Start

Prerequisites

Python 3.11 or higher
pip (Python package manager)

Installation

# Clone the repository
git clone https://github.com/hubertdomagalaa/Machine_Learning.git
cd Machine_Learning

# Create virtual environment
python -m venv venv

# Activate virtual environment
# Windows:
venv\Scripts\activate
# Linux/Mac:
# source venv/bin/activate

# Install dependencies
pip install -r requirements.txt

Running the Cancer Detection API

# Train the model
python scripts/train_cancer_model.py

# Start the API server
uvicorn api.cancer_api:app --reload

# Visit http://localhost:8000/docs for interactive API documentation

Running Tests

# Run all tests with coverage
pytest tests/ --cov=src --cov-report=html

# View coverage report
# Open htmlcov/index.html in your browser

📂 Repository Structure

Machine_Learning/
├── api/                     # FastAPI endpoints for production models
│   ├── cancer_api.py       # Cancer detection REST API
│   └── schemas.py          # Pydantic request/response models
│
├── src/                     # Production Python modules
│   ├── cancer/             # Cancer detection system
│   │   ├── config.py       # Configuration management
│   │   ├── data_loader.py  # Data loading and validation
│   │   ├── preprocessor.py # Feature engineering
│   │   ├── model.py        # Model training and evaluation
│   │   ├── predictor.py    # Prediction interface
│   │   └── cli.py          # Command-line interface
│   └── utils/              # Shared utilities
│
├── tests/                   # Unit and integration tests
│   ├── test_cancer_*.py    # Cancer system tests
│   └── ...
│
├── notebooks/               # Exploratory analysis notebooks
│   ├── Cancer/             # Breast cancer classification
│   ├── Card_Fraud/         # Credit card fraud detection
│   ├── Digits/             # Handwritten digit recognition
│   ├── Flags/              # Country flag analysis
│   ├── Honey/              # Honey production forecasting
│   ├── Medical_Insurance/  # Insurance cost estimation
│   ├── Raisins/            # Raisin variety classification
│   └── income_class/       # Income bracket prediction
│
├── models/                  # Trained model artifacts (.pkl, .joblib)
├── scripts/                 # Training and utility scripts
├── .github/workflows/       # CI/CD automation
├── requirements.txt         # Python dependencies
├── .gitignore              # Git ignore rules
└── LICENSE                 # MIT License

🏥 Cancer Detection System

Problem Statement: Binary classification of breast tumors (Malignant vs. Benign) to assist in early cancer diagnosis.

Business Impact: Early detection significantly improves survival rates. This system achieves high accuracy while minimizing false negatives (missing cancer cases).

📊 Performance Metrics

Accuracy: 96.7%
Recall (Malignant): 100% (No false negatives!)
Precision: 94.2%
F1-Score: 0.97
ROC-AUC: 0.989

🛠️ Technical Stack

Algorithms: Random Forest, SVM, Logistic Regression, KNN
Feature Engineering: Normalization, dimensionality reduction (PCA)
Deployment: FastAPI REST API
Testing: pytest with 85%+ coverage
Data: Wisconsin Breast Cancer Dataset (569 samples, 30 features)

🎯 Key Features

✅ Ensemble voting classifier for robust predictions
✅ Zero false negatives (critical for cancer screening)
✅ Production-ready API with request validation
✅ Comprehensive unit tests
✅ Model versioning and artifact management

🎮 Try Live Demo → | View Project Details → | View Production Code → | API Docs →

💳 Fraud Detection

Problem Statement: Detect fraudulent financial transactions in real-time to prevent monetary losses.

🛠️ Technical Approach

Feature Engineering: Transaction ratios, balance differentials, transaction type encoding
Class Imbalance Handling: SMOTE oversampling, class weights
Model: Logistic Regression (baseline), designed for XGBoost upgrade
Evaluation: Precision-Recall curves, confusion matrix, cost-sensitive metrics

📊 Dataset Characteristics

Highly imbalanced (fraud is rare: <1% of transactions)
Time-series features (transaction steps)
Multiple transaction types (PAYMENT, TRANSFER, CASH_OUT)

View Project Details →

✍️ Handwritten Digit Recognition

Problem Statement: Optical recognition of handwritten digits (0-9) for automated document processing.

🛠️ Technical Approach

Dimensionality Reduction: PCA for visualization and feature compression
Models Compared: SVM, Random Forest, MLPClassifier (Neural Network)
Hyperparameter Tuning: GridSearchCV for optimal parameters
Dataset: UCI ML hand-written digits (1,797 samples, 8x8 images)

📊 Best Model Performance

Algorithm: SVM with RBF kernel
Accuracy: ~98%
Confusion Matrix Analysis: Detailed digit-pair error patterns

View Project Details →

🍯 Honey Production Forecasting

Problem Statement: Predict future honey production trends across U.S. states to assist agricultural planning.

🛠️ Technical Approach

Model: Linear Regression (baseline)
Feature Engineering: Year-over-year percentage change
Data Aggregation: State-level and national trend analysis
Evaluation: MSE, R-squared, residual analysis

📊 Key Insights

Identified declining production trends in key states
Seasonal and economic factor correlations
Multi-year forecasting capabilities

View Project Details →

🏴 World Flags Classification

Problem Statement: Predict country characteristics based on flag features (colors, symbols, patterns).

🛠️ Technical Approach

Data Source: UCI ML Repository (194 countries, 30 features)
Models: Decision Trees, Random Forests, SVM, Neural Networks
Feature Types: Numerical (colors, area) and categorical (symbols, religion, language)
Evaluation: Cross-validation, classification reports

View Project Details →

🍇 Raisin Classification

Problem Statement: Classify raisin varieties using physical measurements for quality control.

🛠️ Technical Approach

Algorithms: Clustering and supervised classification
Features: Size, shape, color characteristics
Application: Automated agricultural sorting

View Project Details →

💰 Income Prediction

Problem Statement: Predict whether individuals earn above or below $50K based on demographic features.

🛠️ Technical Approach

Data: Census income dataset
Features: Age, education, occupation, work hours, marital status
Models: Classification algorithms with feature importance analysis
Evaluation: Accuracy, precision, recall, fairness metrics

View Project Details →

🏥 Medical Insurance Calculator

Problem Statement: Estimate medical insurance costs based on individual health and demographic factors.

🛠️ Technical Approach

Design Pattern: Object-Oriented Programming with Enums
Features: Age, BMI, smoking status, number of children
Validation: Input validation, error handling
Code Quality: Type safety, clean architecture

💡 Software Engineering Highlights

This project showcases professional Python development:

✅ Enum types for type safety
✅ Data validation with custom setters
✅ BMI calculation encapsulation
✅ Comprehensive error handling

View Project Details → | View Code →

🛠️ Technical Skills Demonstrated