Backpropagation Neural Network Ensemble in Predicting Inflation

Abstract

 Global economic volatility that continues to experience spikes is a particular concern for countries in the world, including Indonesia. This is due to the impact that will occur if it continues to increase which can result in a country's economic recession. A country must pay attention to the pressure on the inflation rate. Unreasonable inflation rate volatility can have a negative impact on economic growth. Therefore, it is very important to accurately predict future inflation rates so that it becomes important information for economic policy makers. Inflation prediction is one of the problems that has been widely researched because the data is non-stationary and non-linear, so an algorithm is needed that can overcome this problem. One of the algorithms that can be used is the Backpropagation Neural Network (BPNN), but the BPNN network in its application has many parameters that must be determined so that it often causes overfitting. For this reason, instead of learning from multiple models, the ensemble method is used. The main benefit of this method is to reduce overfitting and at the same time maintain the accuracy and diversity of the BPNN network.