The objective of this research is to identify a gene signature with prognostic significance for predicting the recurrence of colorectal cancer (CRC). The use of conventional risk assessment for patients with stage II/III CRC is still debated. Tumor metastasis is believed to be closely linked to epithelial-mesenchymal transition. Therefore, it is encouraging to develop a predictive model using the epithelial-mesenchymal transition-related gene (ERG) signature of genes involved in the epithelial-mesenchymal transition. A total of 1,780 individuals at stages II and III of colorectal cancer had their transcriptome profiles and clinical data examined retrospectively from 15 publicly available databases. Using a coefficient variant analysis, reference genes for standardizing gene expression levels were chosen. To create the ERG signature for predicting disease-free survival (DFS), univariate, LASSO, and multivariate Cox regression analyses were used. Patients were classified as either high- or low-risk based on their recurrence risk scores. Several CRC cohorts were included in the survival analysis. About 7 cancer-related ERGs and 3 reference genes made up the suggested ERG signature. It was shown that the ERG signature recurrence risk score was a significant prognostic factor in the validation cohort and had a prognostic impact in all cohorts (P<0.05). Both DFS (P=0.0001) and OS (P=0.0058) were lower for high-risk CRC patients in the combined cohort compared to those at low risk. ERG signature showed better colon cancer prediction performance than Oncotype DX. To better assess the prognosis, a decision tree and nomogram were combined. Patients with CRC can benefit from using the found ERG profile as a powerful and promising biomarker for recurrence prediction. The disclosed ERG signature may also aid in patient stratification according to tumor biology and advance the field of personalized medicine.