Concept drift is a term used in the fields of artificial intelligence, big data and digital transformation. It describes a change in the data or patterns recognised by a computer model. This means that data with which a machine or software has been trained changes over time - it "drifts".
Imagine using artificial intelligence that recognises credit card fraud based on previous transaction data. If the behaviour of fraudsters changes, the old data no longer matches the new fraud attempts. At some point, the artificial intelligence recognises new types of fraud less well - its knowledge is therefore "outdated". This process is known as concept drift.
It is therefore important for companies to regularly review their data models and retrain them with fresh, up-to-date data. This is the only way, for example, that a system for detecting fraud, predicting product ends or analysing customer behaviour can remain reliable. Understanding concept drift is essential today in the age of artificial intelligence in order to use digital systems efficiently and securely.