This paper presents the problem of optimizing Radio Access Technology (RAT) selection and traffic offloading in a Multi RAT Heterogeneous Network composed by cellular and WiFi networks. The main idea is to distribute the traffic load of the system over the available access networks, while making an efficient use of the radio resources in the long-term. In this sense, a higher and fair WiFi resource utilization is pursued in comparison to cellular network usage. The goal is to avoid an overloaded cellular network scenario and to obtain a decrease in the energy consumption. The problem is modeled as a sequential decision-making process using Semi Markovian Decision Process, where the optimization problem aims at maximizing the total long-term discounted-reward of the network. Value Iteration algorithm is then used to solve the maximization problem and to obtain the optimal policy. From the resulting policy, several performance metrics are derived to evaluate its performance. Numerical results indicate that the proposed RAT selection strategy performs well for distributing the traffic load in a fair manner among WiFi networks while a lower cellular utilization is achieved.