Forecasting Indonesia Inflation Using Long Short Term Memory Neural Network I Gusti Bagus Ngurah Diksa(a*), Heri Kuswanto (b), Kartika Fithriasari (b)
a) Department of Statistics, Faculty of Science and Data Analytics, Sepuluh November Institute of Technology, Surabaya, Indonesia
b) Department of Statistics, Faculty of Science and Data Analytics, Sepuluh November Institute of Technology, Surabaya, Indonesia
Abstract
The case of inflation can influence monetary policy, and therefore, in assisting policy decision making, inflation forecasts can be made. Inflation forecasting is a connecting bridge to determine the value of inflation for the coming period. Getting an accurate inflation forecast value will be an important thing for determining monetary policy. The running inflation process allows it to change from time to time, resulting in a nonlinear model that will provide a more accurate forecast of inflation. The neural network is a general function approach capable of mapping any nonlinear function. One part of the neural network method that can be used in forecasting is the Long Short Term Memory (LSTM) method. This method has the advantage of storing information for a longer period. The neural network method^s efficiency depends on the network structure of the number of nodes and epochs in converging conditions. This study aims to obtain the best inflation forecasting model in Indonesia using the LSTM method. This method is a development network of a Recurrent Neural Network with a configuration of four gates, namely input gate, input modulation gate, forget gate, and outputs gate. Based on the research results, the best LSTM model in predicting Indonesia^s inflation has more than one node in hidden layer with the optimum number of epochs. However, too many nodes are used, and the not optimal use of epoch will make the root mean square error value of the model even worse.
Keywords: Inflation, Forecasting, Nonlinear, Long Short Term Memory, Root Mean Square Error
