Forecasting Provincial Government Expenditures in Indonesia using Artificial Neural Network

Abstract

The purpose of this research is to analyze using artificial neural network techniques in predicting the realization of provincial government spending in Indonesia according to the type of expenditure with Back-Propagation. This research needs to be done because it can be seen from the side of realization of provincial government spending in Indonesia that there can be a surplus and a deficit. Therefore, it is necessary to make predictions as an effort to address this. The data comes from the publication by the Central Statistical Agency of the provincial government of financial statistics (BPS). Financial provincial government data was collected through local government financial surveys from provincial government agencies in Indonesia. The analysis process uses the help of Rapidminer software and is validated with K-Fold values from 2 to 10. The data is divided into training data and testing data. Training data is data from 2016-2018 and testing data is data from 2017-2019. Several architectural models were tested namely '3-2-1; 3-5-1; 3-10-1; 3-5-10-1 'to obtain an accurate prediction by considering the value of Root Mean Square Error (RMSE). The results of the back-propagation analysis state that the 3-5-1 model is the best model with an RMSE value of 0.027 at k-fold = 9 for training data and an RMSE value of 0.035 for testing data. These results confirm that the back-propagation algorithm can be implemented in this case.