Learning from examples

Choose your learning level

BeginnerIntermediateAdvanced

Start with plain-language explanations. Switch levels at any time.

Beginner explanation

Machine learning finds patterns in examples to make predictions or decisions. Instead of writing every rule yourself, you choose data and a learning procedure. The learned patterns can be wrong or stop working when conditions change.

Intermediate explanation

Features are input measurements; labels are target outcomes when available. Training fits model parameters. Inference applies a fitted model to new inputs. Begin with a baseline, such as predicting the recent average, before trying a complex model.

Advanced explanation

Specify the target, observation time and deployment population before modeling. Leakage occurs when training features include information unavailable at prediction time. Good retrospective fit can be useless when the operational timing is wrong.

PUT IT IN CONTEXT

A worked example

A fictional shop predicts tomorrow’s demand from past sales and known holidays. Tomorrow’s final sales total cannot be an input: it is the value being predicted. A previous-week average is a useful first comparison.

TRY IT YOURSELF

Practice the idea

Choose two valid features and one leaking feature for predicting whether a fictional parcel will arrive late before it ships.

Your working notes (not saved or sent)

Reveal the answer guide

Valid features might include planned route distance and service level known at shipment. Actual delivery timestamp leaks the future outcome and must not be used for this prediction.

Go further: advanced challenge

Describe a data snapshot procedure that guarantees every feature was available at the historical prediction time.

An open-ended discussion task. Explain your assumptions and evidence; there may be more than one defensible answer.

Continue the conversation

What part of this idea would you like to test or question? Compare your reasoning with the community.

Find related forum discussions ↗

Opens a topic search in the existing AstraCruz forum. Results depend on available posts; signing in may be required. No discussion or activity is fabricated.

Browse all forums