🤒 Predicting Influenza 🤒

Developed by 💻:

• Fernando Schettini.
• Augusto Barreto.

Special thanks to 🥰:

• Ricardo Rios, your experience and knowledge have been invaluable to our progress.

About 🤔:

This project analyzes data on suspected COVID-19 Influenza-Like Illness (ILI) cases from Brazil's e-SUS Notifica system in 2024. Using classical AI models, including KNN, Decision Tree, MLP, Perceptron, K-Means Complete Link, and a Bayesian Classifier, the project preprocesses the data and compares the performance of these models.

Resourses 🧑‍🔬:

• Data preprocessing;
• F1 and recall methods comparison;
• Comparative ROC curve graph;
• Silhouette analysis for unsupervised methods;
• Dendrogram for the complete link method;

Results 📈:

The best prediction methods were the decision tree and MLP. Unsupervised methods did not perform well in creating groups of sick and healthy individuals.

SuperVised Models:

Table 1 - Results comparison between methods.

Método	F1-Score (Avg)	Recall (Avg)
K-NN (n=5)	0.7025 ± 0.0087	0.7392 ± 0.0116
Árvore de Decisão (max_depth=5, min_samples_leaf=10)	0.7422 ± 0.0116	0.8359 ± 0.0259
Class. Bayesiano	0.6655 ± 0.2047	0.7392 ± 0.2380
Perceptron (iter=200)	0.5858 ± 0.2347	0.6998 ± 0.3720
MLP (iter=300)	0.7472 ± 0.0135	0.8258 ± 0.0195

Figure 1 - ROC Curve comparison graph.

Non-SuperVised Models:

Método	Jaccard	Silhueta (Avg)
K-Means (n=2)	0.2951239424104527	0.5853294620160377
Complete Link (n=2)	0.31986874200238724	0.48111038632360703

Figure 2 - Silhuette for K-means.

Figure 3 - Silhuette for Complete Link.

Dependencies 🚚:

The project dependencies are described in ./dependencies/requirements. within the repository.

For installing dependencies more quickly, you can run the following command at terminal, inside the clonned repository:

sudo apt update && sudo apt install python3 python3-pip
pip3 install -r ./dependencies/requirements.txt

How to contribute 🫂:

Feel free to create a new branch, fork the project, create a new Issue or make a pull request contact one of us to develop at Predicting Influenza.

Licence 📜:

Apache V2