Introduction to Machine-Learning Practical #2

This is the repository for the 2nd machine-learning practical at the Biotechnology School of Strasbourg (ESBS)
By Corentin Meyer, PhD Student @ CSTB - iCube, 10/01 and 11/02 at the ESBS

Goal of this practical

In this pratical I will try to show you how to create a machine-learning model to predict the diabete risk of patients.
We will perform data exploration, data processing, model training and evaluation and finally fine-tuning. Python is the go-to language for machine-learning and data-science. We will use Jupyter Notebooks through Google Colab to work in this pratical. See the next section called Getting Started to import the pratical into Google Colab.

Getting Started

For this pratical we will sue Google Colab Notebooks. It is online Python notebooks that allow you to code and use python without installing anything on your side. Because it is provided by Google, you will need a Google Account for this to work. If you don't want to use a Google service or you want to learn how you would use python in real life (at a job or insternship) please go to he #Advanced Setup section.

Google Colab usage

1. Go to: https://colab.research.google.com/
2. On the pop-up to select Notebook, select the Github Tab and enter this link: https://github.com/lambda-science/ML-TD-ESBS-2 and select the notebook showing (TD2_Machine_Learning.ipynb).
3. You should be able to run notebook and start writing code ! Create a cell and run print("Hello World") to check if everything is good !

Congratulations you should now be ready to code for the TD ! You should read the second part at least for the informations that it contains ;)

Advanced Setup

Click to open

This is the hard but worth-it way to install Python and all other stuff. It's worth to try as these are the tools you will use for sure in any job or internship. It makes everything run on your computer instead of relying on Google's one. I will briefly explain each tool to you and how to use them.

The tools and how to install them

1. Git
   To Install Git: Git (should already be installed on all Linux) Git is command-line software used in informatics to do code versioning (tracking modifications and updates). In any informatics project you WILL be using it. In this TD we will only use Git to download the TD Code from a GitHub Repository using the command in a terminal:
   git clone https://github.com/lambda-science/ML-TD-ESBS-2.git.
   For people that do not want to use Git and could not install it the code is also availiable as a .zip file HERE

2. Anaconda envrionnement
   Anaconda is a python distribution and package manager. It is the prefered way for data-scientist to install Python. We will use Anaconda to install our python environnement. An environnement is a python installation with a specific set of library installed. This way we can insure that we all have the same package installed with the same version.
   In any Python project, you WILL be using Anaconda environnement (or at least Virtual Environnement). It's is crucial for reproductibility, sharing and tracking.
   To install our environnement
   If not already done, install Anaconda from: Anaconda Download Page
   After cloning the Git repository, you will find a file named environment.yml containing all informations for the Anaconda envrionnement. You can now create the environnement using:
   conda env create -f environment.yml
   Note: If an error occurs such as conda is not a valid command you might need to use the anaconda prompt software for the command. Also environnement are heavy (1.5-2gb here) and can take some time to install.
   You can now activate your environnement in your current terminal using:
   conda activate TD2_ML. Your command-line should now look like (TD2-ML) you@computername:~
   (For WINDOWS please use Anaconda Prompt terminal or run C:\ProgramData\Anaconda3\Scripts\activate base before conda activate TD2_ML if it doesn't work)

3. Jupyter Notebooks
   Jupyter Notebook is the main tool of any data-scientist. It allows you to write and run python code dynamically without reloading all the code, data and variables everytime.
   It is structured as blocks of code that you can run and edit independantly. In this TD, our main worksheet will be the TD2_Machine_Learning.ipynb Jupyter Notebook.
   You have several option to open Jupyter Notebooks.

Using the Tools

• The Easy way: Activate you conda env if not already done. Then run:
  jupyter-notebook
  You browser should automatically open a windows on Jupyter or you can simply click the link. Please check if you can open the ipynb file and the folder.
  Note: Don't close the terminal prompt as it would shutdown the Jupyter Server.
• The 2nd Way: Use VSCode
  Install the python extension. VSCode is able to natively open Jupyter Notebook with a great interface and without a server. Just select the right python environnement and you're ready to go !

Congratulations you should now be ready to code for the TD. Simply open the .ipynb file using jupyter-notebook or VSCode !