A comparison between SARIMA models and Feed Forward Neural Network Ensemble Models for Time Series Data

Abstract

In this paper, we investigate the robustness of Feed Forward Neural Network (FFNN) ensemble models applied to quarterly time series forecasting tasks, by comparing their prediction ability with that of Seasonal Auto-regressive Integrated Moving Average (SARIMA) models. We obtained adequate SARIMA models which required statistical knowledge and considerable effort. On the other hand, FFNN ensemble models were readily constructed from a single FFNN template, and they produced competitive forecasts, at the level of well-constructed SARIMA models. The single template approach for adapting FFNN ensembles to multiple time series datasets can be an economic and sensible alternative if fitting individual models for each time series turns out to be very time consuming. Additionally, FFNN ensembles were able to produce accurate interval estimations, in addition to good point forecasts.

A Literature Review on Some Trends in Artificial Neural Networks for Modeling and Simulation with Time Series

“…For example, predictions can guide people, organizations, and governments to choose the best options or strategies to achieve their goals or solve their problems. Another example of the application of predictions can be found in the consumption of electrical energy to guarantee the optimal operating conditions of an energy network that supplies electrical energy to its customers [1][2][3][4][5]. A time series is a set of records about a phenomenon that is ordered equidistantly with respect to time; this is also called a forecast.…”

“…These models are based on historical values of the time series to extract information about patterns (trend, seasonality, cycle, etc.) that allow the extrapolation of the behavior of the time series [1].…”

“…Another variant of RNN, called nonlinear autoregressive modeling with exogenous inputs (NARX), has feedback connections that enclose several layers of the network, which can be used by including present and lagged values of k exogenous variables x (1) , x (2) , . .…”