| MSc Thesis: Recommending Independent Music to Twitter users - |
The purpose of this project was to identify if implicit data (1.5 million users, 11 million tweets stored in MongoDB) from Twitter can be used to recommend music to users. An open-source recommendation library in Python using several packages such as Divisi2, NumPy, SciPy and PySparse was used to generate the recommendations.
This library contained an implementation of a model-based collaborative filtering approach namely Matrix Factorization using Singular Value Decomposition. The implemented algorithm/library was adopted and improved to recommend relevant music artists to the target user based on their listening history and the other users in the network who had similar preferences.
You may wish to check it out if you have a Twitter account, this was developed using Django and Bootstrap frameworks and deployed on AWS Elastic Beanstalk