The artificial neural network for forecasting foreign exchange rates

A foreign exchange market is one of the highly invested markets in the world with an average daily trade volume of $1.8 trillion. Due to extreme volatility and uncertainty associated with foreign currency fluctuations, the prediction of a foreign exchange rate has been one of the most challenging and onerous tasks for both researchers and practitioners. Traditionally, a foreign exchange forecast has been predicated on some technical indicators that simply tracked past pricing trends without considering a host of other factors (e.g., changes in government policy, trade imbalances, inflation). However, if the currency market is influenced by a random event (i.e., if the market did not follow the trend pattern), these indicators will lead to misleading forecasts. That is to say, traditional forecasting techniques such as a linear trend analysis would not work well for predicting future foreign exchange fluctuations. To overcome this shortcoming of the traditional forecasting techniques, we propose an Artificial Neural Network (ANN) that has proven to be useful for forecasting volatile financial time series such as foreign exchange rates. After applying ANN to the actual data, we discovered that the proposed ANN turned out to be very effective in predicting the daily fluctuations of foreign exchange rates. Similarly, its experiments showed favourable results for weekly forecasts, although it did not perform as well as we anticipated for monthly forecasts.