Next Generation Mobile Networks are envisioned to integrate and coordinate Heterogeneous Networks with the aim to cope with the new mobile traffic demands by taking advantage of the features of each wireless network. In this context, it is accepted that a centralized management resource framework is required by Mobile Network Operators in order to efficiently manage the scarce and limited radio resources from each wireless network. One of the main challenges for radio resource management architecture is the network selection function. We investigate the problem of Radio Access Network selection for a Heterogeneous Networks scenario with the objective of distributing the traffic among several Radio Access Networks in a fair way. The problem is theoretically modeled as a sequential decision-making problem using Semi Markovian Decision Process, where the optimization problem seeks to maximize the long-term discounted reward of the Heterogeneous Network system. In addition to this, taking advantage of the departure of sessions, we have considered a load distribution process that allows offloading of traffic to alleviate the load of the macro cell. In order to solve the model and obtain an optimal policy, we have used Value Iteration algorithm. From the resulting policy, the blocking probability for each possible event in the system is calculated. Several simulations were carried out, and the obtained results indicate that the proposed network selection strategy exhibits good performance for distributing the traffic load in a fair way among several wireless networks.