National Identity Predictive Models for the Real Time Prediction of European School’s Students: Preliminary Results

摘要

An experimental study is conducted to predict the real time national identity (national or immigrants) of the students based on their responses in information and communication technology (ICT) survey held in European schools. All the experiments are conducted in SPSS IBM modeler version 18.1. The target datasets were collected by ESSIE (SMART 2010/0039) during the big survey at levels 3 of schools ISCED (International Standard Classification of Education) in the year 2011. The auto classifier node selected 5 supervised machine learning classifiers filtering out of 8 classifiers. To predict the national identity of students in academic school, the highest accuracy 96.6% is achieved by decision tree C5 with filtering of 46 features out of total 156 and to predict the national identity of students in vocational school, the uppermost accuracy 94.3% is achieved by Tree-AS with reduction of total 41 features out of total 172. Hence, to predict the national identity, self-reduction and auto classifier stabilized only 46 features for C5 Tree and 41 features for Tree-AS. The findings of paper also signify that C5 classifier outperformed the Logistic Regression (LR) and Tree-AS after feature reduction at academic schools. Further, Tree-AS also outperformed the Bayesian network (BN), linear support vector machine (LSVM) and LR after feature reduction at vocational schools.