The latest buzz in the realm of forex robots is neural networks. In essence, neural networks are an ‘artificial intelligence’ computer technology that is designed to interpret data in the same way that the human brain interprets its environment. Unlike traditional data structures neural networks take in multiple data streams and then outputs one result: as long as you can quantify data, you can include it in the factors to get a result – or make a prediction. This is precisely what the expert forex trader does.
However, there is a catch: neural networks need to be taught or trained to recognise and adjust for patterns in the forex market. If a neural network is presented with samples of previous data it should be able to learn the dependencies in the data set – the training set. A testing set is then used to see whether the neural network makes accurate predictions based on the training set, or if adjustments need to be made.
The caveat is this: the neural network will only be able to predict events that would normally fit within the training data series; if a completely outlying event such as the collapse of a big bank occurs the neural network will fail to predict that, and forex robots will cost their owners a substantial amount of money.
Neural networks are becoming an integral part of forex trading platforms and a number of premium forex signals use neural networks to find and recommend trades which forex signal subscribers use on their trading account. Hedge funds, for example, are making heavy use of neural networks in their proprietary high frequency trading systems.
As an individual trader a forex signal based on neural networks could be a good benchmark against which to weigh your trading decisions. However if your only trading strategy is to use a forex robot receiving forex signals generated by neural networks you may be in for a surprise.