An embodiment provides a network parameter optimization method and apparatus, to implement online network parameter optimization. The method includes: sending, by a first function entity, an evaluation model to a second function entity, where the evaluation model is used by the second function entity to determine a first network parameter adjustment action performed by a network element in a first network environment state; receiving, by the first function entity, information about the first network parameter adjustment action; updating, by the first function entity, the evaluation model based on information about the first network environment state, the information about the first network parameter adjustment action, information about a return obtained after the network element performs the first network parameter adjustment action, and information about a second network environment state after the network element performs the first network parameter adjustment action.