Forecasting Inflation Rate in Kermanshah Province with LSTM Neural Network

Document Type : Original Article

Author

Department of Statistics, Faculty of Science, Razi University, Kermanshah, Iran

10.22034/mpo.2023.413712.1092

Abstract

Inflation is a crucial issue whose rate prediction can be helpful in correcting or continuing societal and economic decisions and setting macro policies. Also, forecasting the inflation rate helps investors, businesses, companies, and consumers make informed decisions about the future, ultimately leading to promotion, and economic growth. Therefore, this paper discusses the problem of forecasting the inflation rate in Kermanshah province. Various statistical models have been introduced to forecast time series, including the widely used Seasonal Autoregressive Aggregate Moving Average (SARIMA) model. The purpose of providing them is to provide predictions with high accuracy.
Along with statistical models, machine learning methods, including neural networks, have shown that they are a powerful tool with statistical models in predicting time series. One of the widely used neural networks in this field is the Long-Short-Term Memory (LSTM) network,. According to its architecture, the LSTM network is specifically designed for sequential data, which can accurately remember the long-term dependencies of data. Therefore, in this paper, LSTM neural network is used to predict the inflation rate of Kermanshah province. Using the data of the inflation rate, which was recorded and reported seasonally from March 2013 to June 2023 by the Iranian Statistics Center, the prediction accuracy of the LSTM network has been compared with the SARIMA model. The results show that the LSTM network is competitive with the SARIMA model. Therefore, using the LSTM network and the available data, the inflation rate has been predicted in the coming months after June 2023.

Keywords

Main Subjects