Zansa › Answers › Forecasting and data science
What is a feature in machine learning, in simple terms?
A feature is an input variable the model uses to make a prediction, such as day of week, price, days since last purchase or whether a promotion was running. Choosing and building good features, known as feature engineering, often matters more to results than choosing a fancy algorithm.
For forecasting, typical features include calendar variables, lagged sales (sales a week or year ago), rolling averages, prices, promotion flags and holidays.
One rule is crucial: features must only use information known at the time of prediction. Lagged values need to respect the forecast horizon, because if you forecast four weeks ahead, last week's sales will not be available for the final week of the forecast.
Keep features few, meaningful and documented.
Want this applied to your business? Zansa builds dashboards and forecasts for small and mid-sized companies worldwide. Start with a free intro call.
Book a call