A
user equipment may use a learning model to predict user behavior, or traffic corresponding thereto, and may transmit, to a radio network node, a prediction report indicative of the predicted behavior / traffic. Based on the prediction report the node may proactively schedule predicted downlink resources to facilitate delivery, to the
user equipment, of predicted downlink traffic that may correspond to the indicated predicted behavior / traffic. The
user equipment may indicate that successfully decoded downlink traffic, received from the node according to the scheduled predicted downlink resources, is invalid, or not
usable, by avoiding transmission of HARQ feedback corresponding to the received traffic or by transmitting an invalid scheduled resource indication. The user equipment may analyze a confidence level corresponding to the learning model with respect to a confidence level threshold, which may be dynamically increased by the node in response to invalid traffic, to determine whether to transmit a prediction report.