Background and Objective: College students are particularly vulnerable to mental health problems, including suicidal ideation. The COVID-19 pandemic, and the restrictive measures imposed to limit the spread of the virus, such as lockdown or university closures, have increased the vulnerability of college students to mental health problems. In this particular epidemic context, our objectives were to (1) develop a predictive model of suicidal ideation among college students with a machine learning model; and (2) identify the most predictive factors. Methods: We used random forest models to predict suicidal ideation among 346 French college students involved in the French CONFINS longitudinal cohort. We created models for predicting suicidal ideation at follow-up, based on 128 potential predictors reported at baseline that reflected socio-demographics, health, lifestyle habits, familial characteristics, and COVID-19-related characteristics. Results: The most important predictors identified were depressive symptoms, self-reported mood, and anxiety symptoms. The predictive models showed moderately good mean values for the area under the receiver operating characteristic curve (0.74), sensitivity (0.69), specificity (0.74), and negative predictive value (0.89). To a lesser degree, the level of stress before the COVID-19 pandemic, optimism about the quality of life after lockdown, and health literacy contributed to the suicidal ideation prediction. In a subsample of students that did not report suicidal ideation at baseline, the main predictors were quite similar. Conclusions: Few factors were required for predicting the SI. The strongest predictors were related to the college student’s mental health, and they were not specific to the pandemic. These findings may facilitate the development of a routine screening tool for the early identification of students at risk.