huangy22/NewsRecommender

Combines collaborative filtering with content-based filtering using hierarchical clustering on user networks and Latent Dirichlet Allocation (LDA) topic modeling to address cold-start problems in news recommendation. Ingests Twitter retweet behavior and article metadata from major news publishers to build user communities and infer topic distributions, then ranks fresh articles by cosine similarity to group interest profiles. Includes data scraping utilities and a web interface for Twitter-integrated recommendations.

208 stars. No commits in the last 6 months.