استعمال الخوارزمية الجينية لتدريب الشبكة العصبية الاصطناعية لغرض التنبؤ بأسعار الإغلاق اليومية لمصرف بغداد في سوق العراق للأوراق المالية

Abstract EN

Some phenomena suffer from the presence of chaos and instability in its data as a result of the great development in information technology, which led to the existence of many patterns inherent in that data, and the process of identifying those patterns is considered a major and important problem, given its importance in the prediction process.

So the need arose to know and identify those patterns, and to use these patterns in the prediction process.

On this basis, recourse was made to the use of some modern and flexible technologies in that field, namely The Artificial Neural Networks and the genetic algorithm, and the use of that algorithm to update (train) the weights of The Artificial Neural Network.

Through the results produced by the applied side, and after a set of models of Artificial Neural Networks were built with different displacements in variables and different nodes in the hidden layer.

it was concluded that the best model for the purpose of prediction is the model NN(1, 2, 12; 10).

As that model had the lowest values for statistical measures, Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), and (SBC) Criterion.

In addition, the time taken by that model in the training process is less than the rest of the models, as it reached the number of times that model was trained is (57) repetitions.