Designing a Warning System for Hyperinflation for Iran's Economy Mohammad Hossein PourKazemi

Due to the adverse consequences of inflation in different sectors of the economy, the awareness of the possibility of hyperinflation in the near future makes a great opportunity to avoid the occurrence of hyperinflation. Therefore, this research seeks to design a warning system for hyperinflation. This system using the variables affecting inflation and exploiting the basics of the neural networks appraises the possibility of the occurrence of hyperinflation in the next six months. In this research, from the monthly data of 21 possible variables having effect on inflation from March in 1996 to December in 2011 with the combination of genetics’ algorithm and neural networks the essential variables affected Iran’ inflation are determined. These variables are: Liquidity, government expenditure, labor wage index, interest rates, gross domestic product, Inflation with a lag, and global price index for crude oil. After identifying the fundamental variables using the data related to these variables, the designing of a warning system for inflation is sought. To design such a system, a feed forward neural network with two hidden layers is used. The experimental results of the model indicate the promising performance of the warning system, and the system is able to emit an early warning signal for the occurrence of hyperinflation in near future.