Google Professional Machine Learning Engineer Question 13
Select 3Google Cloud PlatformYou are building a recommendation system for a streaming service that predicts what movies a user is likely to watch next. The data includes user watch history, movie metadata (e.g., genre, cast, director), and user reviews. Which machine learning techniques would be the most appropriate to model user preferences?
- A
Matrix factorization to learn latent features of users and movies for collaborative filtering
- B
Boosted trees to predict user preferences based on movie metadata and reviews
- C
Time-series forecasting to predict the next movie a user will watch
- D
Autoencoders to reduce the dimensionality of movie metadata for feature extraction
- E
Regression to directly predict a user's movie rating
Show answer and explanation
Correct answers: A, B, D
Explanation
For a recommendation system, matrix factorization is a strong choice for collaborative filtering, as it learns latent user and movie features. Boosted trees can handle structured data such as metadata and reviews effectively. Autoencoders may be used to preprocess high-dimensional features like movie metadata for better representation. While regression and time-series forecasting have some applications in machine learning, they are less suited to the specific challenges of this recommendation problem.
- A. Correct.
Matrix factorization is highly effective for collaborative filtering scenarios like recommendation systems, as it learns latent features of users and movies.
- B. Correct.
Boosted trees are suitable for handling structured data like movie metadata and reviews to predict user preferences.
- C. Incorrect.
Time-series forecasting is typically used for temporal data with a sequential dependency, but it is not ideal for this use case as there is no clear time-based sequence for movie recommendations.
- D. Correct.
Autoencoders can be used to reduce the dimensionality of high-dimensional data like movie metadata, improving feature extraction for downstream models.
- E. Incorrect.
Regression is more commonly used for numerical predictions, like predicting ratings directly, but it does not capture complex relationships between users and movies effectively in this context.