A Comparison of Data Mining Methods in Analyzing Educational Data

Abstract

Although data mining has been considered as a silver bullet which magically extracts valuable information from the stacked and unused data, its too many methods frequently confuse and mislead researchers. Therefore, in order to get a satisfying result, researchers need plenty of experience to choose a proper data mining method suitable to the purpose of their research. Unfortunately, in the education field, there are a few studies to point out this problem. In order to resolve this issue, in this paper, a study was conducted to compare Neural Network, Logistic Regression, and Decision Tree on educational data from Korea Youth Panel Survey (KYPS). The result showed the prediction accuracies of the methods were meaningfully different, but it doesn’t mean that the prediction accuracy is the only factor in decision of a specific method. Rather, the result suggested that researchers should consider various aspects of the methods to choose a specific method because each method has its own pros and cons.