Google Professional Machine Learning Engineer Question 14
Select 3Google Cloud PlatformYou are a Machine Learning Engineer tasked with predicting the monthly energy consumption for a large energy provider. The dataset includes 10 years of historical energy usage, weather variables (temperature, humidity, etc.), and customer data (e.g., demographic and usage patterns). The energy provider also wants to identify patterns in customer behavior to better segment their customers. Which of the following techniques should you use to meet these objectives?
- A
Use a time-series regression model to predict monthly energy consumption based on historical energy usage and weather variables.
- B
Apply matrix factorization to predict future energy consumption by identifying latent patterns in customer behavior.
- C
Train a boosted tree model for energy consumption prediction, leveraging weather and demographic variables as features.
- D
Use an autoencoder to identify patterns in customer behavior for segmentation.
- E
Use a time-series forecasting method designed for seasonal data, such as ARIMA, to predict energy consumption.
Show answer and explanation
Correct answers: A, C, D
Explanation
To predict monthly energy consumption, a time-series regression model or a boosted tree model can leverage historical data and additional features like weather and demographics. For customer segmentation, autoencoders can identify patterns in customer behavior. Matrix factorization and ARIMA are less suitable for the given objectives because they either do not directly address the predictive requirements or fail to incorporate additional features.
- A. Correct.
Correct. A time-series regression model can effectively utilize historical energy usage and weather variables to predict monthly energy consumption.
- B. Incorrect.
Incorrect. Matrix factorization is commonly used for recommendation systems or collaborative filtering tasks rather than direct energy consumption prediction.
- C. Correct.
Correct. Boosted tree models, such as XGBoost or LightGBM, can handle complex relationships between features and are effective for regression tasks like energy consumption prediction.
- D. Correct.
Correct. Autoencoders are useful for identifying patterns in high-dimensional data, such as customer behavior, which can assist in segmentation.
- E. Incorrect.
Incorrect. While ARIMA is a commonly used time-series forecasting method, it does not leverage additional features like weather and demographic data, making it less suitable for this task.