Chronic Kidney Disease Detection

In this repository, you can find analysis about the Chronic Kidney Disease dataset you can find in this directory or directly here on the UCI Website. This dataset provide 25 feattures which may predict a patient with chronic kidney disease, this work include differents Machine Learning methods comparison to chose the most accurate solution.

Here the model chosen in the main script is XGBoost, with a macro avg f1 score of ~ 98%

This is the repository architechture :

.
├── Chronic_Kidney_Disease
│   ├── chronic_kidney_disease.arff
│   ├── chronic_kidney_disease.info.txt
│   └── chronic_kidney_disease_full.arff
├── README.md
├── config.json
├── exploration
│   ├── CKD_subtypes.ipynb
│   ├── EDA.ipynb
│   └── models.ipynb
├── main.py
├── requirements.txt
└── trainings
    ├── XXXXXXXXXXX
    │   ├── classification_report.txt
    │   ├── confusion_matrix.png
    │   ├── infos.json
    │   └── model_XXXX.pkl

• 📂 Chronic_Kidney_Disease : The dataset used.
• 📄 config.json : The config file where all the differents settings are stored.
• 📂 exploration : The directory with the differents exploration notebooks :
  • CKD_subtypes.ipynb : Notebook to find potential CKD subtypes with clustering method.
  • EDA.ipynb : The Exploratory Data Analysis of the dataset this notebook aim to find risk factors for CKD from provided features.
  • models.ipynb : The differents models comparisons and feature importance of chosen model.
• 📄 main.py : The main script to launch the Chronic Kidney Disease process. This include data preprocessing, model training and test evaluation.
• 📄 requirements.txt : The python requirement file.
• 📂 trainings : Folder where the main script is going to create folder and files corresponding to a new training. In the created folders we can find :
  • 📄 classification_report.txt : Text file with more evaluation informations about the model.
  • 📄 model_XXXX.pkl : File where the model is stored with metric score in the name.
  • 📄 infos.json : Text file with informations about the the model trained (hyperparameters, score, ...).
  • 📄 confusion_matrix.png : Image showing the confusion matrix.

Instalation

Install Python 3.9.13

Then to install the differents python packages used in the script and notebooks, you can use pip :

cd path/to/repo/directory
pip install -r requirements.txt

Usage

You can read the diferrents notebooks in the exploration directory to discover the dataset analysis and then launch the main script :

python3 main.py

You can also give a custom config to launch the script :

python3 main.py --config-path path/to/your/config.json