Predicting Exam Success Using Machine Learning: An Analysis of Learning Behaviors

The availability of learning activity data from online education platforms has created new opportunities to examine how student behaviors relate to academic outcomes. Within this context, educational data mining has been widely applied to analyze learning patterns and support performance prediction. This paper explores whether students' learning behaviors can be used to predict exam success in a Python learning environment. Exploratory data analysis is used to compare behavioral characteristics between students who passed and those who failed the exam. The prediction task is formulated as a binary classification problem using the variable passed exam. A support vector machine (SVM) classifier is applied to distinguish between pass and fail outcomes, and feature importance analysis is conducted to better understand the contribution of different learning behaviors. The results suggest that engagement-related variables, particularly study time and practice activities, are closely associated with exam success, while demographic features contribute relatively little to prediction performance. These findings are consistent with existing educational data mining research and demonstrate the value of machine learning methods for analyzing learning behavior data.