Document Type : Research Paper
Author
Abstract
Background: The present research is about the psychological characteristics of adolescents and their adjustment levels. Considering the rational and theoretical bases regarding interactive relations between these concepts, a traditional model based on correlation and a new model based on parallel distributed data processing were utilized.
Objectives: The aim of the study was a comparison between the abilities of the mentioned models to predict the levels of adjustment based on the psychological characteristics of adolescents.
Methods: The primary data related to 18 psychological characteristics and 5 adjustment levels were obtained by implementing Persian version of CPI and AISS on 456 male high school students in Tehran. The models of correlation and factor analysis were utilized to extract the optimum combination of factors as predictor components. On these bases a combination of 4 components and 5 independent psychological characteristics with the best proportion of prediction with a capability equal to the original 18 characteristics (α<0.01) were identified. Moreover, according to the numerous effective factors in formation of the psychological characteristics as well as adjustment and complications between their complex and non-linear relationships and interactions, Multilayer Perceptron Artificial Neural Networks (MLPANN) model was also utilized for prediction, and its ability was compared with the Regression model.
Results: The findings showed that for predicting five levels of adjustment, ANNs model has more potential than logistic regression model and if we reduce the number of adjustment to 3 levels, then this capability changes in favour of logistic regression model.
Conclusion: Thus, the particular characteristics of ANNs such as parallel distributed processing and recognition of non-linear and complex relations by learning and experiencing and the special ability of regression model in predicting on the basis of linear relations (prioritization of the role of each predicting factor) is one of the major factors for the success of each model.
Keywords
)