A dynamic hierarchy resource management approach-DHRM based on intelligent prediction was proposed for heterogeneous cognitive network.In DHRM,according to different time scale,the method of wavelet neural network,wiener prediction and reinforcement learning were brought to get the variation of traffic d ion,the resource requirement of the handover calls,and the information of users’preferences,and available hierarchical resources of all networks were allocated flexibly.Multi-attribute decision making method,based on network status and user preference was used to make decision to dynamically assign network traffic flow to the most appropriate network.Simulation results show that,the system capacity is improved about 20% by DHRM compared with the other joint radio resource management algorithms.