Natural factors and non-pharmaceutical interventions (NPIs) have effect on COVID-19 transmission, but it’s difficult to separate these two factors. The Compound natural factor (CNF) model is proposed to deal with this problem. In this model, the weight of single natural factors (SNFs) could be expressed the coupling relationship (CR) among them. Then, CR is iteratively optimized by Elitism-based compact genetic algorithms (ECGAs). Considering optimal coupling relationship of SNFs, CNF has a strong correlation (r=0.56) with the COVID-19 infection rate. However, CNF does not have much correlation with mortality (r=-0.25) and recovery rate (r=-0.46), due to slight change of weather in hospitals. Therefore, a linear weighted CNF model is constructed to forecast the impending infection rate. As a result, NPIs effect have been eliminated in the predicted infection rate by the CNF model, which is only the result of climate change. If China ignored NPIs, COVID-19 virus would transmit in the CNF forecast way as climate changes. This model built on Chinese cases provides a new perspective to forecast the global infection rate which is only under the intervention of natural factors.