charbel-a-hC / ups-ml-stroke-prediction

This repository contains the stroke prediction among patients project included in the UPS (Université Paris-Saclay) ML course curriculum.

Stroke Prediction

In this project, we attempt to classify patients to find out if they will have a stroke or not.

Dataset

This dataset can be obtained from Kaggle and can be found in this link. In the dataset, we have 5110 datapoints (rows) and each row provides the following information on each patient:

• id: unique identifier
• gender: "Male", "Female" or "Other"
• age: age of the patient
• hypertension: 0 if the patient doesn't have hypertension, 1 if the patient has hypertension
• heart_disease: 0 if the patient doesn't have any heart diseases, 1 if the patient has a heart disease
• ever_married: "No" or "Yes"
• work_type: "children", "Govt_jov", "Never_worked", "Private" or "Self-employed"
• Residence_type: "Rural" or "Urban"
• avg_glucose_level: average glucose level in blood
• bmi: body mass index
• smoking_status: "formerly smoked", "never smoked", "smokes" or "Unknown"*
• stroke: 1 if the patient had a stroke or 0 if not *Note: "Unknown" in smoking_status means that the information is unavailable for this patient

Repo Structure

• dataset/healthcare-dataset-stroke-data: full dataset downloaded from kaggle
• Makefile: Makefile where you can create a virtual env or create a docker container to run the notebook
• Dockerfile: Dockerfile to build a docker image with GPU support (tested on NVIDIA TITAN RTX with Ubuntu 18.04). More information on how to download docker can be found here. nvidia-docker installation can also be found here
• StrokePrediction.ipynb: Notebook containing all the code
• poetry.lock/pyproject.toml: Python Poetry files to install a development environment with all dependencies handled
• envrionment.yml: YAML file containing conda exported environment for development (you can use either poetry or conda)

Running the Notebook

Clone and cd in the repository before running any of the commands:

git clone [e-mail]:charbel-a-hC/ups-ml-stroke-prediction.git
cd ups-ml-stroke-prediction

You also need to install python3.8 locally if you wish to run the notebook on a local environment. For Ubuntu:

sudo apt-get install python3.8 \
    python3-pip \
    python3.8-venv \
    python3.8-dev \
    python3.8-distutils

And you need to update your pip:

/usr/bin/python3.8 -m pip install --upgrade pip

Docker

If you have docker installed:

docker build . -t ups-ml-stroke-prediction
docker run -it --rm -v --runtime=nvidia ${PWD}:/ups-ml-stroke-prediction ups-ml-stroke-prediction bash

After launching the container, a notebook will be open in the following ip/port; localhost:8888.

Local Environment (Ubuntu-18.04) - Poetry

Simply run the make command in the repository:

make env

A virtual environment will be created after running the above command. In the same shell, run:

poetry shell

This will activate the environment and then you can open a jupyter notebook:

jupyter notebook --ip 0.0.0.0 --port 8888 --allow-root

Local Environment - Conda

You can download Anaconda here. After the download, open an anaconda navigator prompt if you're on windows and run the following commands:

conda env create -f environment.yml
conda activate ml

Note: If you're on Linux, you can open a normal terminal and run the following command before creating the environment:

conda activate base

You can then open a jupyter notebook similarly:

jupyter notebook

Google Colaboratory

You can open the notebook in Google Colab here: