Machine Learning Workshop (TAA)

Faculty of Engineering, Universidad de la Republica 2024 Course

This repository consolidates the projects developed during the Machine Learning Workshop course, based on the book "Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow" by Aurelien Geron.

Repository Structure

Weekly Workshops

Five hands-on workshops developed throughout the semester, each exploring different machine learning techniques:

• 01-clasificador-titanic - Binary classifier for Titanic survival prediction
• 02-sentiment-imdb - Sentiment analysis on movie reviews
• 03-regresion-bicicletas - Bike demand prediction (Random Forest + XGBoost)
• 04-deteccion-anomalias - Anomaly detection on datasets
• 05-bicicletas-rnn - Time-series prediction with recurrent neural networks

Kaggle Projects

Two competitive projects developed during the course, with full code and documentation:

6. 06-kaggle-higgs-boson - Particle physics event classification (Report)
7. 07-kaggle-freesound - Audio classification with deep learning (Report)

Documentation

The docs/ folder contains PDF reports for all projects:

• higgs-boson-informe.pdf - Project 1: Higgs Boson Challenge
• freesound-informe.pdf - Project 2: Freesound Audio Tagging
• entregable-1.pdf - First project deliverable
• entregable-2.pdf - Second project deliverable

Key Results

• Sentiment Analysis (IMDB): 87% accuracy with TF-IDF, bigrams, and stopword removal
• Bike Demand (XGBoost): RMSLE ~0.37 on cross-validation (Kaggle top 5% ~0.35)
• Titanic Classifier: Identified social class and gender as critical survival variables

Environment Setup

# Create conda environment with all dependencies
conda env create -f environment.yml
conda activate taa

# Launch Jupyter to explore notebooks
jupyter notebook

Tech Stack

• Frameworks: scikit-learn, Keras, TensorFlow, PyTorch
• Data processing: pandas, numpy, matplotlib, seaborn
• Models: Random Forest, XGBoost, RNN/LSTM
• Techniques: TF-IDF, SHAP values, cross-validation