Classifier Model Optimization: Tailoring Machine Learning Models to COVID patients classification

---

Introduction

This project aims to evaluate the accuracy of different machine learning algorithms on the task of determining if COVID patients should be hospitalized or not.

The analysis focuses on choosing the best hyperparameters for each model, using the optuna framework, and comparing each models' accuracies.

---

Configuration

Customize the experiment using the config.ini file. Key parameters include:

Types:

• data_type: Choose img_feature for numerics and images features vectors combined, img for only raw images, and numeric for only the numeric data.
• model_type: Specify dt, nn, cnn or dl for the model type.

---

Running

After the environment for the experiment is built, simply run main.py and check the results on the terminal

---

Analysing Data

Each test will compare different hyperparameters and their accuracies. At the end, the best combination of hyperparameters will be printed on the terminal, with the according result.

Framework

This project was built mainly using the Optuna framework, more information about the intricacies of its functioning can be found here.

---

Disclaimer

This repository is an addition to an academic paper with the same name, for the Machine Learning class at University of Coimbra, Portugal.

The authors are:

• Catarina Silva
• Mariana Guiomar
• Saulo José Mendes