Data-driven Uncertain Modeling and Optimization Approach for Heterogeneous Network Systems

In the age of IoT, the heterogeneous network fusion becomes a tremendous issue, a dilemma in heterogeneous network is how to integrate the resource and allocate the multifarious services which is remarkable but seriously difficult to system-level model and quantify. Aiming at the representation of uncertainly systems design element distribution, combined with modeling and optimization, here we proposed a novel mixture stochastic process and multi combination upper confidence bound strategy for data-driven Bayesian Optimization. This method can be generally applied to the uncertain modeling and design problem in heterogeneous networks scenarios. We applied the method to the multi-services scenario of space information network systems. Compared with other combinations of surrogate models with origin acquisition strategies in the experiments, our method brought up a better representation and optimization results.