This project was concieved in collaboration with Udacity and IBM Watson
In this project I aim to create a recommendation system for the IBM Watson community. The recommendation system suggests articles for users to interact.
The recommender system can make recommendations in a number of ways:
- Collaborative Filtering
- Takes into account the similarity of users and recommends the most popular articles read by similar users
- Rank Based Recommendations
- Recommends the highest ranked articles starting with the most highly ranked
- Content Based Filtering
- Produces recommendations based on similarity to material the user has interacted with previously. Utilizes Natural Language Processing (NLP) methodology to analyse and rank articles by similarity.
- SVD - Matrix Factorization Recommendations
- Utilises matrix operations to predict the ranking (or in this case the boolean interaction variable)
- App Code (Heroku)
- Collaborative Filtering
- Rank based recommendations
- Content based recommendations (user's read articles, or certain specified articles)
- dataset visualisation