Predicting Student Performance in Object Oriented Programming Using Decision Tree : A Case at Kolej Poly-Tech Mara, Kuantan

The paper focuses on prediction of student learning performance in object oriented programming course using data mining technique based on a dataset obtained from Kolej Poly-Tech Mara (KPTM), Kuantan. The objective was to identify and implement the most accurate algorithm for the KPTM dataset and to come up with a good prediction model using decision tree technique. The most relevant rules were identified from the model. The dataset was run through some pre-processing such as data cleaning, data reduction and discretization. The experiments were conducted using machine learning software Weka 3.6.9. The first experiment was to test the clean dataset with seven classification techniques. Accuracy plays an important role to prove the best classification technique by using correctly classified instance as an indicator. Using 10-fold cross validation for each algorithm, it was found that decision tree was the best algorithm with 83.6944% correctness. The second experiment was conducted to find the best model among the percentage split where the best percentage split produced the best model accuracy. The experiment with 50% of data training and 50% of data testing in percentage split produced higher accuracy where the percentage of correctly classified instance was 76.2557%. The rules were extracted from the model and after the analyses were conducted the result showed that the domain factors of student performance were class attendance and the performance of the previous semester