CompTIA DataX Practice Test 2026-2027 and Free Sample Questions

A model shows low error on training data but high error on validation data. This pattern is most consistent with:

Which regularization technique can drive some model coefficients exactly to zero, effectively performing feature selection?

A fraud detection model is evaluated on a dataset where genuine transactions vastly outnumber fraudulent ones. To assess how well the model identifies actual fraud cases while limiting false alarms, the most informative single metric is:

K-fold cross-validation is used during model development primarily to:

A categorical variable named region has the values north, south, east, and west, with no inherent order. The most appropriate encoding for use in a linear model is:

A data scientist models the number of customer support calls received in an hour, where calls arrive independently at a constant average rate. The most appropriate distribution is the:

Random forest improves on a single decision tree primarily by:

In gradient descent optimization, setting the learning rate too high is likely to cause:

A data scientist uses k-means clustering and must choose the number of clusters k. The elbow method helps make this choice by:

When applying feature scaling such as standardization to numeric features, the scaler parameters should be:

An analysis finds a strong positive correlation between ice cream sales and the number of drowning incidents. The most appropriate conclusion is that:

A time series shows a repeating pattern that occurs at the same time each year, such as higher retail sales every December. This component of the series is called:

In a neural network, the purpose of a nonlinear activation function such as ReLU is to:

After deploying a model to production, monitoring for model drift is important because:

Principal component analysis reduces the dimensionality of a dataset by:

A survey is conducted only among visitors to a company's website, and the results are used to draw conclusions about all potential customers. The main threat to validity here is:

In a binary classification confusion matrix, a false negative occurs when the model:

In an A/B test of a new website layout, randomly assigning users to the control and treatment groups is important because it:

SHAP values are used in machine learning primarily to: