Automatic prediction of students in e-learning environments is challenging due to a large amount of education data. This issue is addressed by educational data mining. Education data mining is essential in analyzing and discovering a meaningful pattern in education data. These entire meaningful patterns are helpful in student performance improvement and in making a decision. We proposed a student performance prediction model using data mining techniques in this study. We used an enhanced educational dataset with student behavioural features. These features show the student learner's interactivity with the e-learning system. We applied various classification algorithms to our data set: decision tree, random forest, support vector, and Logistic Regression. We first trained the model using this classification algorithm, then obtained results of the models were compared. The model is evaluated based on performance assessment metrics, Accuracy, precision, and f measure and recall values. The experiment results show that random forest performs better than other tested models. Moreover, we demonstrate different visualization, which is helpful in the interpretation of the result. Furthermore, this study reveals that student behavior features strongly correlate with student performance.