Abstract: Objective: To identify the core genes of endoplasmic reticulum stress (ERS) in chronic obstructive pulmonary disease (COPD) using the bioinformatics, various machine learning algorithms and experimental validation. Methods: The microarray data GSE5058, GSE8545, and GSE19407 were downloaded from the GEO database to identify differentially expressed genes (DEGs) between airway epithelial cells of COPD smokers and non-smokers, and then the common DEGs were obtained after overlapping with ERS-related genes and enriched for analysis. Three machine learning algorithms, LASSO, SVM-RFE, and RF, were used to screen the characteristic genes, and their diagnostic performance was verified and evaluated in the GSE10006. Subsequently, immunoinfiltration analysis was performed. Finally the lncRNA-miRNA-mRNA ceRNA network of key genes was constructed and the mRNA expression levels of lung tissue in the mouse emphysema model were verified. Results: A total of 153 common DEGs were screened, of which 74 genes were up-regulated and 79 genes were down-regulated. GO and KEGG analysis showed that ERs-DEGs were mainly enriched in ERS response, protein folding and multiple inflammatory signaling pathways, and DO analysis was mainly enriched in pulmonary vascular occlusive diseases and COPD and so on. Immunoinfiltration analysis showed that COPD samples were highly correlated with a variety of immune cell infiltrations. A total of 4 characteristic genes (including THBS1, BCL2, USP13 and RNFT2) were finally identified by machine learning algorithms, and they showed good diagnostic performance in both the training set and the validation set. At the same time, co-expressed mRNA and miRNA were selected to construct the mRNA-miRNA interaction network. Reverse transcription-quantitative polymerase chain reaction (RT-qPCR) results showed that compared with the air exposure group mice, the mRNA expression levels of THBS1 and RNFT2 in the lung tissues of emphysema mice induced by cigarette smoke exposure were increased, and the mRNA expression levels of BCL2 and USP13 were decreased (all P＜0.05). Conclusion: THBS1, BCL2, USP13 and RNFT2 may be the core genes formed by ERS during the pathogenesis of COPD, and are expected to be targets for COPD immunotherapy.