NWDAF

NWDAF is the 5G core's built-in analytics brain, standardised by 3GPP in TS 23.288. The idea is to stop every network function from reinventing its own data collection and instead give the core a common function that gathers data — from other NFs, from the OAM system, even from application functions — and hands back analytics that others can act on.

What it actually produces are things like load predictions for a slice, anomaly flags, abnormal-UE behaviour detection, and QoS sustainability forecasts. A consumer NF can subscribe to these and adjust: a slice could be scaled before it congests, or a policy tweaked when a device starts misbehaving. In newer releases NWDAF is split into logical pieces — an Analytics Logical Function and a Model Training Logical Function — which is what makes it a practical home for ML in the core rather than just a glorified counter aggregator. It's a foundational building block for closed-loop automation inside 5G.