In virtual machine(VM) live migration process, the VM selection and host detection which are overloaded or underload is place crucial role. host load is dynamic nature, so detection of overload or under load is very challenging take, the live migration process is done with current host loads state, we proposed method, it use future host state, the future loads detection by using Hidden markov model, by this it avoids intermediate live migration of VM. Our suggested techniques are tested using CloudSim simulations on a variety of PlanetLab actual and random workloads. The experimental results reveal that our suggested algorithms outperform the other competitive algorithms in terms of service-level agreement violations, number of VM migrations, and other metrics.