The client-centric multi-cloud environment is a new trend in cloud computing, because it does not require volunteer interconnection and sharing of infrastructures among cloud providers. Through an important business model, i.e. cloud service broker, cloud application providers are able to offer domain specific services by using cloud resources/services made available by multiple cloud providers in different locations. However, the increasing cloud services offered by different cloud providers at different locations with different prices raise the application placement problem subject to general cost and location aware performance. First of all, current cloud service brokers usually neglect the network latency between cloud users and providers which significantly affects the performance of cloud-based services. Second, the existing studies about service location allocation problems with duplication mechanism take no account of the workflow-based applications. Finally, most existing approaches cannot select the appropriate cloud-based services for static and dynamic requests effectively. This work aims to improve cost efficiency and location-aware performance in multi-cloud environment by novel approaches based on evolutionary computation and reinforcement learning for three typical application placement problems.