You want to assess popularity (mis)calibration of a given recommender system using the popularity calibration metrics %Δ𝜉 ("Delta metrics") we discussed in the lecture.

The items in the interaction history Hu of user u have the following popularity estimates (e.g., total number of users who interacted with the item):

The items in the recommendation list Ru of user u have the following popularity estimates (using the same underlying popularity definition as for Hu):

Please calculate the popularity (mis)calibration of the recommender system for user u in terms of the %Δ𝜉 measure using the median for 𝜉 and reporting the result as percentage (like in the examples provided in the lecture).

You want to assess popularity (mis)calibration of a given recommender system using the popularity calibration metrics %Δ𝜉 ("Delta metrics") we discussed in the lecture.

The items in the interaction history Hu of user u have the following popularity estimates (e.g., total number of users who interacted with the item):

The items in the recommendation list Ru of user u have the following popularity estimates (using the same underlying popularity definition as for Hu):

Please calculate the popularity (mis)calibration of the recommender system for user u in terms of the %Δ𝜉 measure using the arithmetic mean for 𝜉 and reporting the result as percentage (like in the examples provided in the lecture).

Assume you are a developer for a fancy travel platform and just implemented and deployed a new hotel recommender system. It is configured to recommend 5 hotels per user based on their travel style. 

For your first (and only) two users of the system, the lists of recommended items are given below.

What is the item coverage of your recommender? Provide your numeric answer in [0,1].

Assume you are a developer for a fancy travel platform and just implemented and deployed a new hotel recommender system. It is configured to recommend 5 hotels per user based on their travel style. 

For your first two users, the recommender creates the following ranked recommendation lists, which are displayed to each user. You record the hotels (items in the recommendation list) that each user clicks on to display further information and use this data as binary relevance score.

What is the MRR of your recommender? Provide your answer as number in [0,1].

Interpreting the binary relevance scores as gain, what is the mean CG@4 of your recommender, i.e. the mean-averaged CG@4 scores over the two users? Provide your answer as a single number.

Assuming these two users are the only ones you have at the moment, what is the user coverage of your recommender? Provide your numeric answer in [0,1].

Consider the following user-item-rating matrix containing explicit feedback signals on a 7-point Likert-scale [1,7]. Zeros indicate unknown ratings. 

You want to build a memory-based User-KNN collaborative filtering recommender using Pearson correlation coefficient for user-user similarity computation.

Consider the following user-item-rating matrix containing explicit feedback signals on a 7-point Likert-scale [1,7]. Zeros indicate unknown ratings. 

You want to build a memory-based User-KNN collaborative filtering recommender using Pearson correlation coefficient for user-user similarity computation.

Consider the following user-item-rating matrix containing explicit feedback signals on a 7-point Likert-scale [1,7]. Zeros indicate unknown ratings. 

You want to build a memory-based User-KNN collaborative filtering recommender using Pearson correlation coefficient for user-user similarity computation.

F-measure is defined as the geometric mean of precision and recall.