Forecasting Inflation in Indonesia using Long Short Term Memory

Abstract

The problem of inflation has always been a macro problem that occurs in developing countries, especially Indonesia. Observation of the inflation rate cannot be done by observing only certain years because the problem of inflation is a long-term problem. The case of inflation in Indonesia has become quite important to note since Indonesia adopted the inflation target system. Modeling and forecasting the inflation rate is needed and considered important because it is related to poverty alleviation where people with low incomes or still are required to be able to meet their needs with high prices of goods. One of the forecasting models that can be used is the Long Short Term Memory (LSTM). This model is a development of the previous model Recurrent Neural Network (RNN). The results showed that the best model is the model with 5 nodes in hidden layer, Adam optimizer and 0.01 learning rate. Forecasting results show that until September 2022 Indonesia's inflation rate condition will increase but not significantly increase so that it remains stable below 2% until September 2022 and still classified as mild inflation.