What is a validation set used for?
Use it to compare models, tune settings and choose thresholds. Keep the test set separate for a final evaluation.
What is class imbalance in machine learning?
It occurs when one class is much less frequent than another. Review metrics for the rare class and keep its prevalence in mind when interpreting results.
What is a precision-recall curve?
It plots precision against recall as the decision threshold changes. Use it to compare tradeoffs when positive cases matter.
What is feature scaling?
It changes numeric feature ranges, for example by subtracting the training mean and dividing by the training standard deviation. It can help distance-based or gradient-based models.
How do decision trees make predictions?
They route a record through a series of feature tests to a leaf. The leaf supplies the predicted value or class estimate.
What is a random forest?
It combines multiple decision trees trained with randomness in the data or candidate features. Their combined prediction can be more stable than one tree.
What is data drift?
It is a change in the distribution of the data a model receives. Track important features over time and check whether changes affect model performance.
Why document the prediction date in a project?
It defines which information would have been available to the model at the time. That helps reviewers spot features that accidentally use future data.
