استخدام الطرق اللاخطية (Svm، Elm، Mlp)‎ في التنبؤ بالسلاسل الزمنية : دراسة حالة

Abstract EN

This study aimed to use both the Artificial Neural Network (ANN) and the Support Vector Machines (SVM) which considered as a non-linear models, in addition, to use the autoregressive integrated moving average (ARIMA) which represent a traditional model to predict the monthly prices of metal index.

The study used time series data for the monthly prices of metal index during the period of October 1990 to October 2020, therefore the sample size is 361 observations.

Initially, the study determined the best model for each method, results were are follow, ARIMA(1, 1, 0), MLP 5-5-1, ELM 5-100-1 and SVM with (Cost(C) = 1000, Epsilon(ε) = 0.1, gamma(γ) =100).

Finally, the study differentiated between the recommended models by using the predictive accuracy measures such as RMSE, MAE, and MAPE.

Results indicated that ELM model was better than ARIMA, MLP, and SVM, in addition, ARIMA model is better than MLP and SVM.