An Efficiency Comparison of Algorithms and Feature Selection Methods for Predict the Learning Achievement of Vocational Students

Abstract

    The objectives of this research was to compare an efficiency of algorithm and feature selection methods to predict the learning achievement of vocational students. By studying the vocational certificate student’s profiles during the 2007-2016 academic year for 5,100 records 27 features. In this research includes two methods: Simple Classification and Hybrid Classification, Simple classification used classification algorithms only. The three candidate classification algorithms that used in this experiment for both simple and hybrid classification include: 1) Decision Tree: J48graft 2) Naïve Bayes and 3) Rule Induction. Hybrid classification used forward selection techniques for select attributed. The data was analyzed by using Rapid Miner Studio 8 with data mining technique on 10-fold Cross Validation. The comparative results showed that the accuracy of Decision Tree: J48graft by forwarding selection and all feature selection is 83.08% and 81.71% ,respectively. In conclusion, the prediction model using Decision Tree: J48graft is most appropriate for predicting the learning achievement of vocational students