Witnessing the great growth of cloud services, providers are faced with the state monitoring for a diversity of services from federated clouds. Since the consumers might have individual SLA requirements on its applications, the state monitoring of service has become a demanding work in the cloud. In this paper, we present a SLA-driven state monitoring framework for cloud service based on the matrix factorization model, which employs a boosting WCGD algorithm to optimize the monitoring results. Our approach not only contributes to improving the monitoring accuracy, but also takes less overhead during the communication. As demonstrated in our experiment, our framework proves to be efficient and adapts to the dynamical network environment. Compared with existing approaches, we are the first to conduct the state monitoring based on the matrix factorization model.