You work as a software engineer for DisneyPlus, a popular movie streaming service across the globe. The CEO suggests that DisneyPlus provides users with movie recommendations. You are the lead engineer in this project.

$1)$ You start by considering different approaches to build a recommender system. DisneyPlus has many movies in their database. However, they only recently started recording user ratings and have very few ratings in their database. In case you need to implement the system before being able to collect a lot of data, which of the following would you consider to be a better recommender system?

a$)$ User$-$based collaborative filtering

b$)$ Item$-$based collaborative filtering

c$)$ Content$-$based recommendation

Give a brief justification of your answer.

$2)$ A few years later, DisneyPlus has collected lots of data, and the situation in the previous part of the question no longer applies. You are working in the team that develops an algorithm to decide if two songs are similar to each other. To do this for two movies A and B$,$ we compute the Pearson correlation coefficient between the ratings given to movie A$,$ and the ratings given to movie B$.$

Your friend Thomas works at DisneyPlus, and proposes to use a similar strategy to determine if users are similar to each other, based on whether other uses have followed them or not. Whenever a user u follows another user v$,$ Thomas gives the $($user$,$ user$)$ pair a $$rating$$ of r$($u$,$v$)=1.$ Because there are $($millions choose $2)$ pairs of users on DisneyPlus, Thomas does not store any ratings for pairs of users who don$$t follow each other. Thomas again proposes to compute the Pearson correlation coefficient between two sets of ratings.

Is Thomas$$s approach a good strategy? Why or why not?

$3)$ How can you design a recommender system that takes into account the evolving taste of users over time $($i$.$e$.$ newer ratings are more important than older ratings?