Forecasting for Myanmar Currency Exchange Rates By using Back Propagation Neural Network

Abstract

Nowadays, forecasting of exchange rates plays an important role in international economics. Additional to basic economic and financial news, investors and traders employ in their decision process technical tools to analyze the transaction data. In this paper, the system based on neural networks implemented for forecasting Myanmar Currency exchange ratesusing artificial neural network. The system uses back-propagation algorithm to train the exchange rates. Feed forward neural network is used to improve the efficiency of the back-propagation. Multilayer Perceptron (MLP) network is the main architecture. Network architecture parameters especially number of input and number of hidden layers are analyzed.System performance is evaluated in terms of Mean Absolute Error (MAE). Daily historical price data for currency pairs for the last three years are inputted to the system. The system is implemented using programming language C#.