In this paper we present a novel approach for resource allocation in cognitive radio network (CRN) with heterogeneous user traffic. In this approach we deploy some form of reinforcement learning, and make a short-term resource allocation based on the long-term traffic prediction. The corresponding resource allocation algorithm derived in the paper is implemented in cognitive 3rd Generation Partnership Project Long Term Evolution (3GPP LTE) network. Performance analysis of the algorithm has shown that the proposed approach for resource allocation achieves better performance than other schemes designed to deal with the problem of heterogeneous user applications.