Recommender Systems with Machine Learning - Offline Evaluation Techniques

The video tutorial discusses four key aspects of evaluation: task, dataset, metrics, and data partitioning. It explains how tasks can be evaluated using rating predictions and top-N recommendations. The tutorial also covers data representation using a User-Item Rating Matrix (URM) and the importance of data partitioning in recommendation systems. The focus is on understanding how to evaluate recommendations effectively by considering relevant and non-relevant ratings.

Which of the following is NOT one of the four key aspects of offline evaluation?

• Task

• Data Partitioning

• Dataset

• User Feedback

In rating prediction, what indicates a good recommendation?

• A predicted rating significantly lower than the true value

• A predicted rating close to the true value

• A predicted rating significantly higher than the true value

• A predicted rating of zero

What is the primary goal of top-N recommendations?

• To ignore user preferences

• To recommend all available items

• To select the top-N items based on relevancy

• To rank items by price

How is a dataset typically represented in recommendation systems?

• As a URM (User-Item Rating Matrix)

• As a set of user feedback comments

• As a list of user names

• As a collection of item descriptions

What does the 'ground truth' in a URM refer to?

• Predicted ratings

• Zero ratings only

• Non-zero ratings that are known

• All possible ratings

What are the two classifications of ratings in a dataset?

• Relevant and Non-Relevant

• Known and Unknown

• Positive and Negative

• High and Low

What is the purpose of data partitioning in evaluation?

• To delete irrelevant data

• To separate data into training and testing sets

• To increase the size of the dataset

• To combine all data into one set