Praful932/Kitabe

Implements a hybrid recommendation approach combining FunkSVD (gradient-based matrix factorization) for collaborative filtering with TF-IDF term frequency matching on book metadata, trained on the goodbooks-10k dataset of 6M ratings across 10k books. The Django web application preprocesses book-rating pairs to handle duplicates and missing data, then uses learned book embeddings to detect patterns like author similarity and genre inference. User ratings are processed through both the embedding matrix and token-based matching to surface complementary recommendations where each method excels.

181 stars. No commits in the last 6 months.