Databricks Machine Learning Associate Question 620
Single answerYou are training a machine learning model in Databricks to predict customer churn. During your experiments, you notice that when you use a linear regression model, the predictions are inaccurate on the training data, but when you switch to a highly complex deep learning model, the predictions on the test data are poor. Based on the bias-variance tradeoff, what is the most likely cause of this behavior?
- A
The linear regression model has high bias, while the deep learning model has high variance.
- B
The linear regression model has high variance, while the deep learning model has high bias.
- C
Both the linear regression model and the deep learning model have high bias.
- D
Both the linear regression model and the deep learning model have high variance.
Show answer and explanation
Correct answer: A
Explanation
The scenario describes a classic example of the bias-variance tradeoff. The linear regression model, being too simple, cannot capture the complexity of the data, leading to high bias and poor performance on the training data. On the other hand, the deep learning model, being highly complex, overfits the training data and fails to generalize to the test data, which indicates high variance.
- A. Correct.
The linear regression model is too simple and underfits the data, leading to high bias. The deep learning model is overly complex and overfits the training data, leading to high variance.
- B. Incorrect.
This is incorrect because high variance in the linear regression model would indicate it overfits the training data, which is not consistent with the scenario described.
- C. Incorrect.
This is incorrect because the deep learning model in the scenario overfits the training data, which is a result of high variance rather than high bias.
- D. Incorrect.
This is incorrect because both models do not exhibit high variance. The linear regression model underfits due to high bias, and the deep learning model overfits due to high variance.