https://github.com/nickthegroot/recipe-recommendation

:hamburger: Repo for "Personalized Recipe Recommendation Using Heterogeneous Graphs" (2023)
https://github.com/nickthegroot/recipe-recommendation

Last synced: 28 days ago
JSON representation

:hamburger: Repo for "Personalized Recipe Recommendation Using Heterogeneous Graphs" (2023)

• Host: GitHub
• URL: https://github.com/nickthegroot/recipe-recommendation
• Owner: nickthegroot
• Created: 2022-11-28T23:26:52.000Z (about 2 years ago)
• Default Branch: main
• Last Pushed: 2024-01-07T03:38:34.000Z (about 1 year ago)
• Last Synced: 2024-10-27T22:52:58.343Z (3 months ago)
• Language: Jupyter Notebook
• Homepage: https://recipe.nickthegroot.com
• Size: 40.7 MB
• Stars: 1
• Watchers: 2
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        


   

   


Personalized Recipe Recommendation Using Heterogeneous Graphs



**Authors**:

- Nicholas DeGroot (Halıcıoğlu Data Science Institute, UC San Diego)

## Description

This project was created for UCSD's DSC 180: Data Science Capstone. According to the university, the course:

> Span(s) the entire lifecycle, including assessing the problem, learning domain knowledge, collecting/cleaning data, creating a model, addressing ethical issues, designing the system, analyzing the output, and presenting the results.

>

> https://catalog.ucsd.edu/courses/DSC.html#dsc180b

## Getting Started

This project is configured with `devcontainer` support. This automatically creates a fully isolated environment with all required dependencies installed.

The easiest way to get started with `devcontainers` is through [GitHub Codespaces](https://github.com/features/codespaces).

1. Click [here](https://github.com/codespaces/new?hide_repo_select=true&ref=main&repo=571806935) to create a new codespace on this repository.

   - Alternatively, this can be done through the `gh` CLI.

2. Configure the codespace to your liking. We recommend the 8-core machine.

3. Start the codespace and connect. It might take a minute to install all the dependencies. Grab a :coffee:!

4. Connect to the codespace through your preferred method (browser / VS Code).

## Testing

This project is setup with an array of tests using `pytest` to ensure things are working. With a working environment, run the following command.

```

make test

```

### Testing on DSLMP

For UCSD students & staff, we've ensured that everything works on the Data Science Machine Learning Platform servers.

The (auto!) published Docker image contains everything you need to test the project. Under the hood, it's running the same container that any `devcontainer` is.

In DSMLP: log in with your credentials, then run the following:

```

launch.sh -s -i ghcr.io/nickthegroot/recipe-recommendation:main

cd /app

make test

```

This will begin a full run of every test in the project. Currently, this includes a full pipeline test and a smaller data processing test.

## Downloading/Preparing the Data

1. Download the data by creating an Kaggle account and downloading the [`shuyangli94/food-com-recipes-and-user-interactions`](https://www.kaggle.com/datasets/shuyangli94/food-com-recipes-and-user-interactions) dataset.

2. Unzip the data into `data/raw`.

   - You should see a number of files, including `data/raw/RAW_interactions.csv` and `data/raw/RAW_recipes.csv`

3. Run `make data` to clean the data into its cleaned form.

## Running

All models can be trained using `python src/cli/train.py`.

- Run `python src/cli/train.py --help` for all configuration options

- In general, all models can be trained via `python src/cli/train.py --model {model}`

  - For example, `LightGCN` is trained with `python src/cli/train.py --model LightGCN`