In this paper we model access selection in heterogeneous access network environments as a binary linear optimization problem. The model is based on a flexible utility function, which allows the integration of various parameters of merit. One of the major advantages of our proposal is that it considers both the willingness of the end-users to have connectivity just in particular moments (according to the corresponding service patterns), as well as the historical evolution of such services (outcome of earlier problem instances). We assess the feasibility of the proposed method with an illustrative network deployment, embracing both heterogeneous technologies and two services, with different requirements and characteristics. The results show that an appropriate selection of the utility function is of outer relevance to clearly understand the goodness of the obtained performance. © 2013 IEEE.