Question 1.

choose one commercial recommender and describe how you think it works (content-based, collaborative filtering, etc). Does the technique deliver a good experience or are the recommendations off-target?

IMDB: FAQ for IMDb Ratings

According to FAQ for IMDb rating, IMDB uses hybrid recommendation systems that combines content based and collaborative modelling and provides an explanation for increased user acceptance.

Collaborative filtering This method uses inter-user comparisons to generate new recommendations.
A collaborative system con- sists of a database which contains the users’ ratings and is augmented as the user interacts with the system over time.
Advantage
- Works for any kind of item
Disadvantage
- cold start : need enought users in the system to find a match
- Sparsity: hard to find users that have rated the same items
- First rater: Cannot recomment an item that has not been previously rated
- Popularity bias

Content-Based Filtering Content based filtering uses item-to-item correlation to compare rep- resentations of content in an item to representations of content in items the user has rated. The similarities between items and the rating information are used to predict how much the user will like or dislike a new item.
Advantage
- No need for data on other user
- Able to recoomend to users with unique tates
- Able to recommend to new/unpopular items
- Able to provide explanations
Disadvantage
- Finding the appropriate features is hard
- Overspecialization

IMDB uses combined recommended system feature like below. It analyzes of rating from user, also utilize database with other users preference, and this comibation of different methods provide recommended movies to users.

Question 2.

Attacks on Recommender System Travis M. Andrews, The Washington Post (2017): Wisdom of the crowd?

Can you think of a similar example where a collective effort to alter the workings of content recommendations have been successful? How would you design a system to prevent this kind of abuse?

By using fake user profiles into the rating database in order to bias the system’s output, collaborative Filtering based Recommender Systems are the most sensitive systems to attacks. This approache of technique is to create fake user profiles and issue high or low rating to the “target item”. Example of attacks on recommender system would be some recommender system based on reviews and rating like product’s review. False review from fake users can hi-jack the recommender system and gives fales information to users.