Cloud computing responds to user demands with providing computing and storage services. Most of cloud providers, suggest two payment plans to users, namely reservation and on-demand. In reservation plan users have to reserve resources and pay to cloud providers in advance despite the uncertainty in regards to demand. but in on-demand plan, users only pay in exchange for resource usage. Commonly, price of resources in on-demand plan is more expensive than that is in reservation plan. In this paper, we propose a cost effective cloud resource provisioning with imperialist competitive algorithm optimization (CECRPICAO). After predicting demand with a demand predictor algorithm, CECRPICAO allocate the virtual machines with reservation and on-demand plans using these predicted demands. The performance of proposed algorithm is evaluated by numerical studies and simulation. Results show that compared with several prior resource provisioning methods, CECRPICAO can provide the more affordable solution.