INTRODUCTION: Keloids are a common complication that occurs after injury. The pathogenesis of this disease remains unknown. Therefore, identifying immune-related biomarkers and macrophage polarization types in keloids can provide new insights into their treatment. METHODS: In this study, keloid-related bulk RNA-seq data (GSE83286, GSE212954, GSE92566, and GSE90051) were obtained from the Gene Expression Omnibus (GEO) database. The datasets GSE83286, GSE212964, and GSE92566 were combined to form a training set, while GSE90051 was utilized as an external validation set. Differentially expressed genes (DEGs) were detected by comparing keloid and normal samples within the training set. Differentially expressed immune-related genes (DIRGs) were then determined by intersecting the DEGs with immune-related genes (IRGs). Based on the protein-protein interaction (PPI) network, the top 40 DIRGs were selected for further analyses. Weighted Gene Co-expression Network Analysis (WGCNA), in conjunction with three machine learning techniques - least absolute shrinkage and selection operator (LASSO), support vector machine-recursive feature elimination (SVM-RFE), and random forest (RF) - employed to identify biomarkers. Subsequently, a nomogram model was constructed and validated. Single-cell RNA (scRNA) analysis was used to examine the expression of biomarkers at the cell-type level. Furthermore, since keloid is a chronic inflammatory disease and the abnormal polarization of macrophages is essential for the occurrence of this kind of disease, in this study we also endeavor to elucidate the state of macrophage polarization dysregulation within keloid, with the anticipation of generating novel concepts for the treatment of keloid. Finally, western blot (WB) and immunofluorescence (IF) analyses were carried out to confirm the expression levels of the biomarkers. RESULTS: A total of 740 DEGs were identified in the training set, comprising 331 up-regulated genes and 409 down-regulated genes. After intersecting with the IRGs, 73 DIRGs were obtained. Subsequently, the top 40 DIRGs were chosen for further analysis. Eventually, two biomarkers, namely BMP1 and IL1R1, were identified through WGCNA and the three machine learning methods. Their expression levels were then verified by single-cell analysis, WB, and IF analysis. Additionally, it was found that the number of M2 macrophages significantly increased, while the number of M1 macrophages decreased in keloids compared to normal samples. CONCLUSION: BMP1 and IL1R1 might function as novel biomarkers and potential therapeutic targets for keloid treatment. Moreover, upregulating M1 macrophages and downregulating M2 macrophages could represent a promising approach for the treatment of keloids.