The COVID-19 preparedness plans by the Centers for Disease Control and Prevention strongly underscores the need for efficient and effective testing strategies. This, in turn, calls upon the design and development of statistical sampling and testing of COVID-19 strategies. However, the evaluation of operational details requires a detailed representation of human behaviors in epidemic simulation models. Traditional epidemic simulations are mainly based upon system dynamic models, which use differential equations to study macro-level and aggregated behaviors of population subgroups. As such, individual behaviors (e.g., personal protection, commute conditions, social patterns) can't be adequately modeled and tracked for the evaluation of health policies and action strategies. Therefore, this paper presents a network-based simulation model to optimize COVID-19 testing strategies for effective identifications of virus carriers in a spatial area. Specifically, we design a data-driven risk scoring system for statistical sampling and testing of COVID-19. This system collects real-time data from simulated networked behaviors of individuals in the spatial network to support decision-making during the virus spread process. Experimental results showed that this framework has superior performance in optimizing COVID-19 testing decisions and effectively identifying virus carriers from the population.