An Investigation of Data Mining Classification Methods in Classifying Students According to 2018 PISA Reading Scores

Abstract

The purpose of this research was to determine classification accuracy of the factors affecting the success of students' reading skills based on PISA 2018 data by using Artificial Neural Networks, Decision Trees, K-Nearest Neighbor, and Naive Bayes data mining classification methods and to examine the general characteristics of success groups. In the research, 6890 student surveys of PISA 2018 were used. Firstly, missing data were examined and completed. Secondly, 24 index variables thought to affect the success of students' reading skills were determined by examining the related literature, PISA 2018 Technical Report, and PISA 2018 data. Thirdly, considering the sub-classification problem, the students were scaled in two categories as “Successful” and “Unsuccessful” according to the scores of PISA 2018 reading skills achievement test. Statistical analysis was conducted with SPSS MODELER program. At the end of the research, it was determined that Decision Trees C5.0 algorithm had the highest classification rate with 89.6%, the QUEST algorithm had the lowest classification rate with 75%, and four clusters were obtained proportionally close to each other in Two-Step Clustering analysis method to examine the general characteristics according to the success scores. It can be said that the data sets are suitable for clustering since the Silhouette Coefficient, which is calculated as 0.1 in clustering analyses, is greater than 0. It can be concluded that according to achievement scores, all data mining methods can be used to classify students since these models make accurate classification beyond chance.