Introduction PROS Budi W, Suparti, Moch AM Performance of neural fulltext

Proceedings of the IConSSE FSM SWCU 2015, pp. MA.103–108 ISBN: 978-602-1047-21-7 SWUP MA.103 Performance of neural network model in forecasting Indonesian inflation Budi Warsito , Suparti, Moch Abdul Mukid Department of Statistics, Diponegoro University, Semarang, Indonesia Abstract This paper evaluates the usability of neural network for inflation forecasting. The capturing of nonlinear relationships among inflation and its determinants is the base of using this method. The performance of neural networks is quantified by MSE, in both in- sample and out-of-sample prediction. The optimal architecture is specified through a looping process, based on the minimum out-of-sample performance. A simple specification and specialized estimation procedures seems to play significant roles in the success of the neural network model. The data analysis shows that neural network’s forecast of the Indonesian inflation give a significant improvement in forecasting accuracy from some other models analyzed in this paper. Keywords inflation, performance, neural network, forecasting

1. Introduction

The fundamental processes that produce economic series such as inflation, social gaps, or jobless are potentially enough complicated Monge, 2009. This condition makes it very difficult to forecast them. Empirical analysis is largely based on times series. A time series can be either linear or nonlinear. The simple linear approaches have been tried to model them. However, the investigation of nonlinearities in time series is important to macroeconomic theory as well as forecasting. This paper evaluates the usability of neural network for inflation forecasting. The capturing of nonlinear relationships among inflation and its determinants is the base of using this method. Neural network modeling has been gaining attention as an attractive technique for estimation and forecasting in economics Monge, 2009. The prominent advantage of the neural network methods is that they are free from the assumption of linearity that is commonly determined in order to make the traditional methods tractable. Most of the applications have been evolved in financial statistics. This paper appraise whether the nonlinear methodology of ANN significantly improves upon inflation forecasts obtained by traditional linear methods. The proposed method is applied in the case of Indonesian inflation. For the comparison, three other models are also analyzed: subset ARIMA, Kernel Gauss and 2nd Order Spline Suparti et al., 2015. In each model, the best model is selected based on mean square errors of out-of-sample prediction. Neural network is a computational paradigm modeled on the human brain that has become popular in recent years. It perform a variety of tasks, including function approximation, pattern classification, clustering, and prediction. Neural network is very powerful when fitting models to data Samarasinghe, 2006. It can fit arbitrarily complex Corresponding author. Tel.: +62 812 269 1363; E-mail address: budiwrst2gmail.com Performance of neural network model in forecasting Indonesian inflation SWUP MA.104 nonlinear model or nonlinear function to any desired accuracy. Neural network is also capable of time series forecasting, in which the next outcome or outcome for the next several time steps are predicted. This is accomplished by capturing temporal patterns in the data in the form of past memory, which is embedded in the model. In forecasting this knowledge about the past defines future behaviour. How to select these weights and how to obtain optimal architecture are key issues in the use of neural network. The usual approach is to derive a special-purpose weight selection algorithm. In this note, we investigate the usefulness of a neural network model for forecasting inflation. The optimization training is done based on Levenberq-Marquardt method. The performance of neural network is quantified by MSE, in both in-sample prediction and out-of-sample prediction. The selected lags as input variables are obtained from the best subset ARIMA model. The optimal architecture, i.e the optimal number of hidden neuron in hidden layer, is specified through a looping process, based on the minimum out-of-sample performance. The reached of maximum epoch and the achieved of minimum MSE, were the rules of thomb of the stopping criteria.

2. Materials and methods