Classification of Scale Items with Exploratory Graph Analysis and Machine Learning Methods

In exploratory factor analysis, although the researchers decide which items belong to factors by considering statistical results, decisions taken sometimes can be subjective in case of having with similar loadings and complex structures. The aim this study was examine validity classifying into dimensions graph analysis (EGA), has been used determining number recent years machine learning methods. A Monte Carlo simulation performed a total 96 conditions including average loadings, sample size, per dimension, dimensions, distribution data. Percent correct Kappa concordance values were evaluation When findings obtained for different evaluated together, it seen that methods gave results comparable those EGA. Machine showed high performance terms percent values, especially small medium-sized samples. all where loading .70, BayesNet, Naive Bayes, RandomForest, RseslibKnn accurate classification performances above 80% like EGA method. Simple Logistic RBFNetwork also demonstrated acceptable or under many conditions. general, supported these results. revealed whether across is done accurately not.