nickthegroot/recipe-recommendation

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

Hey! I'm Nick, aka @nickthegroot 👋

• 👨‍💻 Work: Senior Software Engineer for Digs
• 🎒 Education: B.S. Data Science, B.S. Cognitive Science (spec. ML & Neural Computation) from UCSD
• 💬 Contact: [email protected]
• 🔗 Website: https://nickthegroot.com

Commonly Used Languages/Frameworks

recipe-recommendation's Issues

Website Publishing

Publishing of #11 to the internet. Ideally automated via GitHub Actions or Netlifly.

Data Reformat

Converting dataset into PyG format for use in models.

Simple GNN Model

Creation of a simple GNN model in PyG. Ideally should attempt to implement an existing recommendation system GNN architecture model for simplicity.

Data Validation

Verifying used variables have format expected (non-null, type, etc.)

Baseline Model

Baseline non-gnn model for comparison. Ideally, this can be done entirely with a third-party library. Check CSE 158 lecture notes for potential libraries/code.

Website Creation

Converting the writeup content from #10 to an actual website. The framework used is up to the developer. Some potential frameworks include:

• Next.JS
• Jekyll

User Persona Creation

Creation of 3 unique "user personas", each with their own favorite/disliked items. These should be codified into an example analysis dataset with item interactions.

Should be tied to some fictitious (but real-sounding) name to help explainability.

Data Visualization

Plotting/graphing of dataset to view variable statistics

User Persona Analysis

Running user personas through all models and analyzing the outputted recommendations.

Data Download

Documented and/or automated via make command.

Analysis Writeup

Simple writeup in wysiwyg editor about user persona analysis results. This will later be converted into a website, so emphasis is on the actual content rather then the visual style.

Complex GNN Model

Creation of a custom GNN architecture for recipe/user interactions