- Feature extraction - creates a profile of the user (describes the types of items the user likes).
- Calculates the items that may be recommended (tf–idf vectoriser tokenises documents, creates a vocabulary of the most frequently occurred words and returns the most relevant items as recommendation). Content-based recommenders use CountVectorizer(), TfIdfVectorizer, cosine similarity.
Recommends the most popular items(top-rated item by the most number of users (trending list)).
Top 10 Most Voted
Top 10 weighted average IMDB
Top In Category
Top Retail Products
Recommends items based on users’ past behavior.
- User-Based similarity
- Item-Based similarity Uses Pearson's Corellation, KNN(k-Nearest Neighbor), SVD(Singular Value Decomposition)