مقارنة الشبكات العصبية الاصطناعية و تحليل الانحدار المتعدد في التنبؤ ببعض المتغيرات النفسية لدى عينة من طالبات الجامعة

Abstract EN

This study aimed to reveal the efficiency and effectiveness of the predictive model built using artificial neural networks in predicting psychological variables among female students from Qassim University comparing to multiple regression analysis with sample size reached 359 students.

The study tools were academic adaptation inventory as a dependent variable, the psychological hardness questionnaire and the self-efficacy scale as independent variables.

Results assured the efficiency of the prediction model that was created using ANN, where the sum of the square errors was used to judge the quality and improvement of prediction between the training data set, where the indicator reached 92.139, and the test set, where the indicator reached 34.042, and it clearly shows a decrease in the error rate for prediction between the two groups, which indicates the validity of the used model and its best predictive performance.

In order to judge the effectiveness of the model, a comparison was made between artificial neural networks (ANN) and multiple linear regression (MLR).

The results showed that there was a decrease in the values of all judgment indicators (root mean square error RMSE - mean absolute error MAE - Mean Absolute Percentage Error MAPE - Root Mean Square Percentage Error RMSPE) and an increase in the explanatory ratio of variance for the predicted variable calculated by means of the coefficient of determination R2, and the network reached a high accuracy rate and gave a solid model that is suitable for predicting future values using the cross-validation method.