What is a feature in machine learning?
A feature is an input variable used by the model, such as the number of purchases a customer made before a prediction date.
What is a target variable?
The target is the outcome the model learns to predict, such as delivery time or whether a customer renews.
Why encode categories for a model?
Some algorithms require numeric inputs. One-hot encoding represents each category with a separate indicator without implying a numeric order.
How can I handle missing data before training?
Inspect why it is missing. If you use an imputer, learn its replacement values from the training data and apply the same rule elsewhere.
When should I use stratified sampling for classification?
Use it when you want random splits to preserve class proportions. Keep group or time boundaries intact when those constraints matter more.
What is mean absolute error?
It is the average absolute difference between predictions and actual values. It uses the same units as the target.
What is root mean squared error?
It is the square root of the average squared prediction error. Squaring gives large errors more influence.
What is model calibration?
Calibration checks whether predicted probabilities agree with observed frequencies. Among many cases scored near 70 percent, a calibrated model should be right about 70 percent of the time.
What is a machine learning pipeline?
A pipeline combines steps such as preprocessing and fitting into one workflow so you can apply the same learned transformations consistently.
How do I perform error analysis on a classifier?
Review false positives and false negatives by relevant groups or conditions. Look for recurring patterns and verify changes on held-out data.
See the beginner Q&A guide for related topics.
