Empirical Evaluation of Hybrid Time Series Forecasting Method Between ARIMA and RBFNN Under Parallel Model

Commonly, a time series is a series of values observed over time. This type of data is used in almost every field ranging from science, engineering, business, finance, economics, medicine to government. The time series prediction problem is the problem of using a model that analyzes past data to predict future values. Accurately forecast time series data is crucial to support decision making in many areas of life. Therefore, improving the accuracy of this problem is a challenging task for researchers in this field. Many traditional time series models or artificial neural network models have been used for time series prediction. Among them, Auto Regressive Integrated Moving Average (ARIMA) model and Radial Basis Function Neural Network (RBFNN) has been widely studied and used for forecasting in specific fields such as finance, stock prices, air pollutants, etc. In this study, we introduce the method of combining two models ARIMA and RBFNN in parallel. This hybrid method can combine the advantages of ARIMA and RBFNN models to create a more effective model for the problem of forecasting time series data. The parallel hybrid method between ARIMA and RBFNN is experimentally evaluated on on real time series data from different fields. The evaluation is based on two criteria: execution time and accuracy. Experimental results demonstrate the power of the proposed method on real-time series datasets in terms of accuracy when compared with each model used individually, but the execution time is a bit longer.