Formation of neural network models for time series data with intervention and its application in CPI data

Abstract

The purpose of this study is to get an optimal forecasting model of CPI data in Yogyakarta using neuralnetwork models that is built with the involvement of the intervention effect. This case happens because of the policyof rising fuel prices. CPI education, recreation, and sport have a pattern like a step function, so that the exact RNNmodels are the models with the input as the truncated polynomial spline regression models. The results of modelpredictions both training and testing data showed high accuracy.

Key words: neural network, intervention, CPI