Time Series Forecasting Using a Hybrid ARIMA and ANN Models

By Malik Arif Arrahman

Student of S1 Statistics

Muhammadiyah University of Semarang

The journal, "Time Series Forecasting using a Hybrid ARIMA and ANN Models", by G. Peter Zhang seek to forecast the time series with hybrid methodology. In this paper, a hybrid methodology that combines both ARIMA and ANN models is proposed to take advantage of the unique strength of ARIMA and ANN models in linear and nonlinear modeling.

The problem in this research comes from the following perspectives. 

First, it is often difficult in practice to determine whether a time series under study is generated from a linear or nonlinear underlying process or whether one particular method is more evective than the other in out-of-sample forecasting. 

Second, real-world time series are rarely pure linear or nonlinear. They often contain both linear and nonlinear patterns. 

Third, it is almost universally agreed in the forecasting literature that no single method is best in every situation.

The focus of research in this journal is about the method to be used not to the dataset. So the discussion in this journal is more emphasized to the method, a case of explanation of methods, disadvantages, and advantages when the two methods are combined.

In this paper, authors propose to take a combining approach to time series forecasting. The linear ARIMA model and the nonlinear ANN model are used jointly, aiming to capture different forms of relationship in the time series data. The hybrid model takes advantage of the unique strength of ARIMA and ANN in linear and nonlinear modeling. 

For complex problems that have both linear and nonlinear correlation structures, the combination method can be an evective way to improve forecasting performance. 

The empirical results with three real data sets clearly suggest that the hybrid model is ableto out perform each component model used in isolation. 

Three well-known data sets---the Wolf's sunspot data, the Canadian lynx data, and the British pound=US dollar exchange rate data---are used in this study to demonstrate the efectiveness of the hybrid method. The sunspot data we consider contains the annual number of sunspots from 1700 to 1987, giving a total of 288 observations.

The method used by the author in this study is appropriate according to the focus of the research whose aim is to combine the two forecasting methods. But the author does not explain clearly the steps of analysis of each method, but only explained in general terms.

Based on the author's presentation in this research, the results show that for short-term forecasting (1 month), both neural network and hybrid models are much better in accuracy than the random walk model. 

For longer time horizons, the ANN model gives a comparable performance to the ARIMA model. The hybrid model out performs both ARIMA and ANN models consistently across three diferent time horizons and with both error measures although the improvement for longer horizonsis not very impressive.

In the concluding section contains conclusions, but in this journal the authors only conclude by method, while for the results of the study were not included in the conclusions. The author also does not include suggestions as hopes for better research conditions, as one solution to the problems that occur.

Overall writing systematic has been arranged well and clearly starting from the research title, author's name, abstract, introduction, method, results and discussion, conclusions, and bibliography. The research title used by the author is quite clear, accurate, and unambiguous, and illustrates what will be examined.