This study aimed to develop a prediction model to classify students based on their academic procrastination tendencies, which were measured and classified as low and high using a self-report tool developed based on the students' assignment submission behaviours logged in the learning management system's database. The students' temporal learning traces were used to extract the features used in the prediction models. The study participants were 51 students enrolled in the Database Management Systems course, which was conducted online using the Moodle learning management system. The study compared the performance of different machine learning algorithms in predicting students' academic procrastination tendencies, analysed the important features of prediction models, and examined whether there is a difference between the academic performance of low and high academic procrastinators. Logistic regression was found to outperform other classification algorithms and reached 90% accuracy in classifying low and high academic procrastinators. Students' regular and early access to course activities were found to be important features in predicting their academic procrastination tendencies. In terms of academic performance, the findings support the existing literature. Students with low academic procrastination tendencies got significantly higher final grades than those with high academic procrastination tendencies. These findings show that students' academic procrastination tendencies can be predicted with high accuracy using online learning trajectories. Such a model will be important in the development of intervention methods for preventing academic procrastination.