Financial Time Series Forecasting Using Hybridized Support Vector Machines and ARIMA Models

(1)Conventionally, time series predicting problems are being solved by several single models. This study presented an innovative hybrid method to combine machine learning model (SVM) with time series model based on autoregressive moving average (ARIMA). The objective of this paper is to solve limitations of ARIMA model in financial time series forecasting by combing it with SVM model. Proposed combination method hybridized both ARIMA and SVM models to capitalize on the unique strength of ARIMA and SVM models in linear and nonlinear modeling. The real exchange rate between Sudanese Pound and Euro (SDG-EURO) used as the data set. Experimental outcomes indicate that the combined model can be an efficient way to increase forecasting accuracy accomplished by either of the models utilized individually. The proposed hybrid model arithmetical outcome examined against benchmark models and around cases in the related literature. The comparison proved the superiority result over all singular alternative models, depends on RMSE error's measure for model accuracy.