🧠 ML Assignments – PPGI • UFPB | 2025

A modular repository for machine learning exercises and experiments. Developed as part of the Machine Learning discipline (MSc in Informatics – PPGI/UFPB).

---

📂 Project Structure

ML/
├── configs/              # YAML configuration files
├── notebooks/            # Jupyter notebooks for each task
├── outputs/              # Generated results and plots
├── src/                  # Core implementation
│   ├── ml/
│   │   ├── data.py       # Dataset generation utilities
│   │   ├── viz.py        # Visualization tools
│   │   └── metrics.py    # Evaluation and metrics functions
│   └── tasks/            # Specific ML algorithms (KNN, SVM, ANN, etc.)
├── .gitignore
├── pyproject.toml
├── uv.lock
└── README.md

---

⚙️ Setup

uv sync

---

📊 Results Overview

Data Configuration (GLOBAL)

• Train/Test split: test_size = 0.20, random_state = 42, stratify = y

✅ Supervised Models

Model	Notebook	Accuracy	Precision	Recall	F1
⭐️ Artificial Neural Network (ANN)	ann.ipynb	1.0000	1.0000	1.0000	1.0000
Logistic Regression	logreg.ipynb	0.9833	0.9677	1.0000	0.9836
SVM	svm.ipynb	0.9833	0.9677	1.0000	0.9836
K-NN (best k = 3)	knn.ipynb	0.9833	0.9677	1.0000	0.9836
Decision Tree	tree.ipynb	0.9500	0.9655	0.9333	0.9492

Notes: Metrics refer to the test split (20%). Each notebook includes the confusion matrix and decision boundary visualization.

---

🌀 Unsupervised Models

Model	Notebook	Primary Metric(s)	Value
K-means	kmeans.ipynb	Adjusted Rand Index (ARI)	0.9342
Fuzzy C-Means	fcm.ipynb	Centers / Iterations	Centers ≈ [[60.07, 30.89], [24.51, 9.80]]; Iter = 15

Notes: K-means includes the confusion matrix and decision region visualization.
FCM includes fuzzy membership heatmaps.

---

🧩 Features

• Synthetic dataset generation — bivariate Gaussian distributions
• Modular architecture for reproducible ML tasks

---

👨‍💻 Author

Lucas G. Dantas
MSc Informatics – AI (PPGI/UFPB) • R&D Data Scientist — Computer Vision | GenAI

---

© 2025 Lucas G. Dantas — All rights reserved.