Objective To establish an efficacy prediction model for immunotherapy of non-small cell lung cancer (NSCLC) by screening the important immune indexes in the tumor microenvironment on the base of immunohistochemical technique.Methods A retrospective analysis was made on 46 patients with NSCLC conformed by postoperative pathology in our hospital between 2013 and 2015. Patients were divided into groups by cluster analysis, and their pathological samples and related clinical data were collected. The immunohistochemical indexes were screened by cluster analysis and t test for dimensionality reduction. A prediction model of NSCLC immunotherapy efficacy was established, and named according to the density and location of immune cell infiltration. The predictive effectiveness of the model was evaluated by validation in 11 cases of NSCLC patients who had immune checkpoint inhibitors treatment.Results The 46 patients were divided into two groups by cluster analysis, 8 patients in group 1 and 38 in group 2. The median disease-free survival (mDFS) of patients was 14.33 months in group 1 and 25.84 months in group 2 (P=0.015). The median overall survival (mOS) of patients were respectively 16.00 months and 28.70 months in the two groups (P=0.021). The dimension was reduced from 16 immune indexes to 7 by cluster analysis and t test. Using these indexes, the prediction model of immunotherapy efficacy was established according to the Fisher discriminant function. The immune cells in the two groups were further analyzed and separately named as immune low response type and immune high response type. The evaluation results showed that the model had a good prediction value.Conclusion The prediction model of NSCLC immunotherapy efficacy based on tumor microenvironment provides a new method for evaluating the efficacy of NSCLC immunotherapy.