Alcohol-withdrawal seizures (AWS) are an important and relevant complication during detoxification in alcohol-dependent patients. Therefore, it is important to evaluate the individual risk for AWS. We apply a random forest algorithm to assess possible predictive markers in a large sample of 200 alcohol-dependent patients undergoing alcohol withdrawal. This analysis showed that the combination of homocysteine, prolactin, blood alcohol concentration on admission, number of preceding withdrawals, age and the number of cigarettes smoked may successfully predict AWS. In conclusion, the results of this analysis allow for origination of further research, which should include additional biological and psychosocial parameters as well as consumption behaviour.