Goal

To make a proof of concept implementation of Tik Tok's collaborative-filtering algorithm for X product category. I'm thinking that food dishes and clothing are good options. The frontend client can be a simple React-based web app that allows users to swipe left or right on recipes (indicating likes/dislikes). The data will serve as training data for the algorithm. The algorithm will improve over time as more users provide more feedback.

Sample Dataset

• https://www.kaggle.com/pes12017000148/food-ingredients-and-recipe-dataset-with-images

Algorithm

• collaborative filtering
  • https://medium.com/swlh/an-easy-guide-to-creating-a-tiktok-like-algorithm-3e9c954fb4e9
  • https://en.wikipedia.org/wiki/Collaborative_filtering#Model-based
  • https://newsroom.tiktok.com/en-us/how-tiktok-recommends-videos-for-you/
  • https://www.datacamp.com/community/tutorials/recommender-systems-python

Database

• raw data → AWS s3 bucket
• SQL Database (PostgreSQL)
  • items table
  • users table
  • user engagement table (generated training data)

Frontend

• React
  • log-in and log-out feature
  • allow user to see their liked items
  • tinder style yes/no flicker
• jupiter notebook python code that implements collaborative filtering on fake users, inputs.
• how long does it take to train the model on larger and larger amounts of inputs?
• how does the model measure performance?
• sort items by most popular and present them to the user
• select random items and present them to the user
• predict what a user would like and present the to the user