Globally, colorectal cancer is the most common cancer, and its incidence and mortality rates are significantly and rapidly increasing. In Taiwan, colorectal cancer screening and chemotherapy have improved the survival rates of patients but also increased the number of patients with recurrence. This study aimed to use machine learning to identify the risk factors for and clinical characteristics of colorectal cancer recurrence. From 2011 to 2018, 2,490 datasets, including 830 from patients with recurrence, were collected; 11 predictive variables were considered risk factors for recurrence based on literature review and physician discussion. The machine learning techniques employed included support vector machine, linear discriminant analysis, logistic regression, C4.5 decision tree, and random forest. SpreadSubsample sampling method was used to resolve imbalance problems. The results showed that the risk factors for recurrence, as ranked by importance, were pathological stage and surgical margins of the primary site. The C4.5 method had the highest mean predicted accuracy (0.8763). In patients who received chemotherapy, pathological stage group should be followed up for those with the colon cases and the number of invaded regional lymph nodes should be considered in patients with the rectum site. Clinical characteristic findings showed that for patients who did not receive chemotherapy, surgical margins of the primary site and number of invaded regional lymph nodes were risk factors that required joint follow-up for different primary sites.