Improve resource provisioning for applications hosted in the cloud is a major research challenge. This is because is necessary to address two conflicting objectives: meet customer requirements; and save money. As there is a delay, which can take minutes, between the request for a new resource and it be ready for use, it is necessary to predict the future demand for each time interval. However, there is not a prediction model that is appropriate in all cases. To resolve this problem, it is proposed in this paper, the combination of different forecasting models by the analytic hierarchy process. In this way, we intend to create a generic solution, able to optimize the allocation of resources to several applications types, with different demand types. The results obtained by simulation show that our proposal achieves this goal.