Pıaac 2015 Türkiye Performansının Yapay Sinir Ağı ve Regresyon Analizi Yöntemleri ile İncelenmesi

Abstract

In this study, the data of the “Program for the International Assessment of Adult Competencies - (PIAAC)” carried out by the OECD to evaluate adult skills were examined by regression analysis (LR) and artificial neural network (ANN) model. Scores of adults included in the PIAAC study in the Turkish sample were predicted by both analysis methods. Demographic features obtained from the background questionnaire within the scope of PIAAC were determined as predictive features in our study. The literacy skill test, numerical skill test and PS-TRE test performances of the adults were accepted as the skill scores that we wanted to predict. Binary logistic regression is used for LR analysis, while multilayer perceptrons are used for ANN model. The effects of independent variables on adult skill performance were determined for both methods. The "IBM SPSS v.23" package program and The IEA International Database Analyzer was used in the analysis of the data. When the results obtained were examined, the rate of correct estimation of literacy skills scores was found to be 69.8% with both analyzes, and 73.6% with LR analysis for numerical skills scores, and 73.5% with ANN. While the rates of estimating PS-TRE skill scores correctly were 67.1% with LR analysis, it was found as 67.3% with ANN analysis. Both analysis methods showed the same performance in classifying literacy skill scores. LR is 0.01% more successful than artificial neural network in classifying numerical skill scores. ANN is 0.02% more successful than LR analysis in classifying PS-TRE skill scores.