Inflation forecasting by hybrid singular spectrum analysis – multilayer perceptrons neural network method, case of Indonesia

Inflation is one of the most important macroeconomic indicators which affects the economic condition of a nation. Therefore, it is necessary to maintain its stability in order that it will not lead to a negative impact and an economic vulnerability. The drastic change in the rate of inflation is determined by the condition of the price of goods which is affected by the distribution and supply-demand factors of goods. As a consequence, it becomes a very important act of action to control inflation. This can be achieved by meeting the information needs of future inflation rates that is needed for the government and the policy of the monetary authority. Fulfillment of accurate and reliable future forecasts of future inflation estimates can be obtained through forecasting. This paper examines the application of the method of Hybrid singular spectrum analysis - a multilayer perceptions neural network to predict the inflation. The main data source used is monthly inflation (in percent) collected by BPS Statistics Indonesia. The result of the study found that the ability of SSA-MPNN Hybrid method is good enough in predicting monthly inflation, as it is provided by the MAPE value of 35.42 percent, without-sample of three observations.