This is a repository demonstrating the bulid of a recommendation system using data directly from dynamodb.
The process consists 3 major steps
- Load and transform json format data from dynamodb to user-item-rating format dataframe
- Use transformed data to train the recommendation engine with lightFM and save the model as a pickle file
- Create a flask API to call the finished model so it can be used on web service
- If new user, items with high average ratings for be recommended
- If old user, items with be recommended using collabrotive filtering