Recommender Systems
Types of Recommender Systems Algorithm
Collaborative filtering algorithm
recommend items to you based on ratings of users who gave similar ratings as you.
- Collaborative does not need the features of the items to be given
Limitations
- cold start problems:
when you have a new item, there are few users have rated, or we have a new user that's rated very few items, the results of collaborative filtering for that item or for that user may not be very accurate. - side information not used:
it doesn't give you a natural way to use side information or additional information about items or users.
Content-based filtering algorithm
recommend items to you based on the features of users and features of the items to find a good match.
Recommending from a large catalogue
- two-step procedures
- retrieval:
- generate large list of plausible item candidates
- combine retrieved items and remove duplicates
- A larger retrieval list gives the ranking system more options to choose from which should maintain or improve recommendations.
- A larger retrieval list may take longer to process which may increase response time.
- ranking
- take list retrieved and rank using learned model
- display ranked items to user
- retrieval:
Similarity Metrics
- The features of an item are basically transformed into vectors/matrixes, and similarity between them can be measured.
- Similarity metrics are often used in content-based filtering.
- There are several types of similarity metrics
| Metrics Type | What is computed | Often for which algorithm | Often for what data |
|---|---|---|---|
| Cosine Similarity (Distance Function#Cosine similarity) | the cosine angle between the vectors | content-based filtering | sparse and high-dimensional |
| Euclidian Distance (Distance Function#Euclidian distance) | the elementwise squared distance between two vectors | content-based filtering | dense and low-dimensional |
| Pearson correlation coefficient | coefficient of the two vectors | collaborative filtering | dense and low-dimensional |
| Dot Product | cosine angle and magnitude of the vectors | collaborative filtering | dense and low-dimensional |