AI-300 Practice Question 94: Data Quality Signal

Microsoft Machine Learning Operations Engineer Associate (AI-300) · Test 10 · question 4 of 10

Implement machine learning model lifecycle and operationsmedium

An upstream system change causes some input features to arrive as null, with the wrong data type, or outside the values seen in training. The team wants the model monitor to detect exactly this kind of problem. Which signal should they add?

Select one answer
Show answer and explanation

Answer:

  • D. The data quality signal

The data quality signal tracks the integrity of a model's input data by comparing it with training data or recent production data. It checks null values, type mismatches and out-of-bounds values, and reports the null value rate, data type error rate and out-of-bounds rate (option D).

- A is wrong because prediction drift compares the distribution of the model's outputs. It does not check the integrity of the input values. - B is wrong because data drift measures shifts in the statistical distribution of input values, not null values, type mismatches or out-of-range values. - C is wrong because this signal tracks changes in how much each feature contributes to predictions, so it does not check input integrity.

Reference: Model monitoring in Azure Machine Learning

Practice Test 10 ›

Study guides for AI-300