دراسة مقارنة للتنبؤ بمعدلات الرطوبة النسبية (RH%) لمحافظة نينوى باستخدام نماذج السلاسل الزمنية الموسمية والشبكات العصبية الاصطناعية

Abstract

The neural network is one of the most important fields of artificial intelligence، which reflects an important development in a significant way of human thinking، and spin around the idea of neural networks simulate the human mind using the computer. Foreseeable development in this area may be due to several studies، which has in the field of neural processing simulation by solving the problems that face، and by following the self-learning processes that rely on the expertise stored in the network، which yield better results. It is intended models of seasonal time series is a set of viewing values associated with each generated sequentially with the continuation of time and refer to replicate the pattern of movement of the time series in the opposite months during successive years. Research has found that the best model is the use of neural networks (NN (1،12; 10)) which includes (12) variable backward in time and (10) hidden contract. For the time-series models seasonal has been the model SARMA (1،0،1)، (0،1،1)₁₂. we have been using some statistical measurements to test the advantage two modes in the prediction which was its result is the style of the time series as seasonal got the lowest values for tests (RMSE،MAE،MAPE،MAD) respectively. It was the best way to predict the relative humidity (RH%) of the province of Mosul is using the model SARMA (1،0،1)، (0،1،1)₁₂