INST0060 Group Project - Group E

This is the code archive by Group E for the INST0060 Group Project.

Aim of this Project

This project aim to explore the use of logistic regression model to compare between two countries' average daily COVID-19 deaths. We particularly focus on whether grouping countries by Continents or by Income Group can help to improve classification accuracy.

How to Use

Install Dependencies

The best practice recommended is to create an environment for the project and install dependencies. To do this, navigate to the root directory of the project, create a new environment, and install dependencies via requirements.txt. With conda as an example:

Create a new environment

conda create -n covidcomp python=3.7

Activate the new environment

conda activate covidcomp

Install dependencies from requirements.txt

conda install -n covidcomp --file requirements.txt

Note: the requirements.txt is adapted from the provided requirement file in INST0060 module. seaborn is added to the requirements for plotting a heat map of correlation matrix.

Run the experiment

To conduct the experiment, the OWID COVID-19 Dataset needs to be provided.

Start the experiment with main.py using command-line arguments. For instance:

python3 main.py ./covid.csv

Prompts of experiment running status will be printed in the console. Plots will be saved to ./output/

Playground

A playground jupyter notebook ./playground.ipynb has been provided alongside with the experiment script to enable interactive experimenting.

Project Structure

• ./covidcomp/ is the main Python library implemented for the project.

  • covidcomp.data module processes the data from the original CSV file to Raw Representation and Derived Representation.

  • covidcomp.model module provides an ABC Model, with is inherited by all concrete model implementation. LogisticsRegression and L2RegularisedLogisticRegression are implemented.

  • covidcomp.experiment module provides ExperimentRunner and ExperimentResult to run experiments on partitioned and flat datasets with cross-validation and store results as ExperimentResult instances.

  • covidcomp.plot module provides a Plotter that will plot the required figures in the experiment.

• ./fomlads/ is the supporting Python library provided in INST0060 Foundation of Machine Learning, from which covidcomp is developed.

• ./data/ contains all the auxiliary datasets.

• ./output/ is the output directory for plots.

• ./playground.ipynb see above.

• ./main.py the entry to the programme, see above.

• The rest are project related configurations and IDE settings

How to Contribute

Code Linting

Code linting is done before committing your code through pre-commit, configured with .pre-commit-config.yaml. In order to enable the pre-commit linting hooks, you will need to install pre-commit in your environment and install the hooks.

Install pre-commit (using conda as an example):

conda install pre-commit

Check if pre-commit is installed properly:

pre-commit --version

Install hooks as specified in the config file:

pre-commit install