The development of science popularity personnel is an important symbol measuring all countries’ level of science popularity. In this paper，we study science assessments on talent development the country’s 31 provinces，based on existing studies to construct a science talent assessment model. The model science talent evaluation algorithm consists ofbinomial regression model based on response surface method （RSM）. Its main feature is that when evaluated based on the evaluation of large public data，each attributes weight information is a completely unknown to program evaluation studies，and thevalidation of this model highlight the advantages of high reproducibility. Applying of this model to the development of science talent inthe country’s 31 provinces to make an assessment，the assessment results show that the model is feasible and effective to build.