Objective To explore biomarkers for the diagnosis of Parkinson disease （PD） and their correlation with immune infiltration based on bioinformatics and machine learning algorithms. Methods The GSE20164， GSE20314， GSE20333， and GSE24378 datasets from the Gene Expression Omnibus （GEO） were selected for analysis to screen for differentially expressed genes in the substantia nigra of PD patients and healthy controls. Gene Ontology （GO） enrichment analysis， Kyoto Encyclopedia of Genes and Genomes （KEGG） pathway enrichment analysis， LASSO Logistic regression algorithm， and random forest algorithm were used to screen hub genes， and the area under the receiver operating characteristic （ROC） curve （AUC） of hub genes for diagnosing PD was calculated. CIBERSORTx was used to evaluate the infiltration characteristics of 22 immune cells in PD patients. Results A total of 20 differentially expressed genes related to PD were screened， including 5 upregulated genes and 15 downregulated genes. GO enrichment analysis and KEGG pathway enrichment analysis showed that 20 differentially expressed genes were involved in dopamine biosynthesis， amine biosynthesis， toxin response， tyrosine metabolism， dopaminergic synapses， PD， synaptic vesicle circulation， and other aspects. LASSO Logistic regression algorithm and random forest algorithm screened out three diagnostic hub genes， KCNMB3， SDC1， and EPYC. The ROC curve analysis showed that the AUC for the comprehensive diagnosis of PD by the three hub genes was 0.783. Immune infiltration analysis showed that the proportion of immature B cells and monocytes in the PD group was higher than that in the healthy control group， and the difference was statistically significant （P＜ 0.05）. There is a positive correlation between immature NK cells and activated CD4+ T cells， and the difference was statistically significant （P＜0.05）. Conclusions The KCNMB3， SDC1， and EPYC hub genes screened through LASSO algorithm and random forest algorithm show good performance in the diagnosis of PD.