feature
/ FEE-chur /
A feature is a single measurable piece of information that describes something you want a machine to learn about. If you are trying to predict the price of a house, the features might be its floor area, the number of bedrooms, the year it was built, and how far it is from the nearest train station. Each feature is one column in a table; each house is one row. The machine never sees the house itself, only this stack of numbers and categories that stand in for it.
Choosing the right features is often more important than choosing the algorithm. A weak feature is one that has little to do with the answer — the colour of the front door tells you almost nothing about price. A strong feature carries real signal. Sometimes the most useful feature is not measured directly but built from raw data: from a birth date you can compute age; from a timestamp you can pull out the day of the week. That craft of inventing better features is called feature engineering.
Be honest about the limits. A model can only learn from the features you give it — if a crucial fact is missing from the columns, no amount of clever maths will recover it. And features can quietly carry bias: if a feature is a stand-in for race, gender, or postcode, the model may discriminate without anyone intending it to. The features are the lens through which the machine sees the world, and a narrow or warped lens produces a narrow or warped view.
To predict whether a loan will be repaid, useful features might be income, the size of the loan, and how reliably past bills were paid. The applicant's name is a useless feature; their debt-to-income ratio — built by dividing two other features — is often a strong one.
Some features carry signal, some carry none, and some are best built from others.
Don't confuse a feature (an input the model reads) with a label (the answer it is trying to predict). The same fact can be a feature in one task and a label in another: a customer's age might be an input for predicting spending, but the label if you are trying to guess age from a photo.