On-Time Graduation Prediction System Using Data Mining Classification Method

The collection of data on academic information system database of Higher Education is often not utilized maximally, whereas from data with data mining technique can give knowledge which not yet known before. The purpose of this research is to know how to form the prediction model of student's graduation rate on- time at Buddhi Dharma University of Tangerang through student passing data. Prediction of student graduation on-time using comparison of algorithm C4.5 and K-NN done with data selection stage, data transformation, data mining and interpretation. This study uses 300 training data and 90 data testing. Then the process of classification technique using decision tree method using C4.5 algorithm and Euclidean distance calculation using K-NN algorithm. Evaluation of classification performance is done to know how well the accuracy of a model is formed. Based on the research that has been done, the model is formed with the help of Rapid miner software, and calculated average value of k-fold cross validation on testing up to k = 10 for algorithm C4.5 and K-NN. Testing is done with Confusion Matrix and ROC curves. Accuracy results obtained prove that Algorithm C4.5 yields 90% accuracy percentage and K-NN yield 87% accuracy percentage. Thus the C4.5 algorithm has a higher accuracy value than K-NN. This C4.5 algorithm can be used as prototype predictions of students' graduation on-time at Buddhi Dharma University Tangerang.