Dropping out of school is a problem in most countries worldwide and leads to many adverse effects on the family and society. Therefore, early predicting the risk of dropping out of high school can help educators have interventions and efficient solutions in reducing the high school dropout rate. In this work, we present a machine learning model to predict the risk of school dropout. This model was built from the dataset, including 10,219 student records with 807 dropouts (7.89%) in high schools in Ca Mau province. The results show that the models Naïve Bayes, Decision Tree with Bagging, Random Forest with Bagging give the best results with Area Under the Curve at 83.01%, 80.95%, 83.16%, and accuracy, precision, recall, f1-score are all over 80%. In addition, we also extracted important features playing a decisive contribution in predicting school dropout, including Grade Point Average, school code, Conduct, Age, and Class.