The Examination Of Forecasting Power Of Econometrics And ANN Models Of Inflation In Iran

Inflation is one of the most fundamental economic problems in each country, so inflation’s trend forecasting for arranging economical policies is very important. This necessity has caused serious attention to the application of different models for forecasting inflation’s rate, thus different forecasting models have developed in competition with one another. Hence this study aimed to forecast the monthly inflation rate in Iran in 1390, has performed using monthly time series data from Iran’s consumer price index of goods and services in the years 1383 to 1389. Information about consumer price index of goods and services for desired years has been obtained from Central Bank of Iran. Hence this study has used two models, Auto Regressive Integrated Moving Average (ARIMA) and Artificial Neural Network (ANN), and has compared the forecasting power of neural network models and econometric models by taking a mean absolute percentage error. Forecast results using these two models showed, although both ARIMA and ANN according to percentage of absolute prediction error within the sample, respectively, 0.86 and 0.94, have a high forecasting power, but ARIMA’s model, in comparison with ANN’s model, has higher forecasting power. Therefore in this study predicted value of consumer price index of goods and services in Iran is determined based on ARIMA time series model. Forecasts show, due to the growing trend in the consumer price index of goods and services in Iran in 1390, choosing monetary policies and liquidity management through appropriate fiscal and monetary policies by policy makers, play an important rate in controlling inflation.