The new trend in Forex system software design is using artificial neural networks for prediction. In particular, prediction of time series using a multi-layer feed-forward forex neural networks. In order to understand how this new forex software works we need to understand how these new forex systems treat prediction.

First, the topic of prediction will be described together with classification of prediction into types. After that, the prediction using neural networks (NNs) will be described. The focus will be on the creation of a training set from a time series. Currency trading software uses Java applets that create a training set and that show the result of a prediction using a neural network of back-propagation type both on functions and financial data.

Predicting is making claims about something that will take place, often based on information from past and from current state.

Everyone solves the problem of prediction every day with various degrees of success. For example weather, harvest, energy consumption, movements of forex (foreign exchange) currency pairs or of shares of stocks, earthquakes, and a lot of other stuff needs to be predicted.

In technical analysis predictable parameters of a system can be often be expressed and evaluated using equations - prediction is then simply evaluation or solution of such equations. However, practically we face problems where such a description would be too complicated or not possible at all. In addition, the solution by this method could be very complicated computationally, and sometimes we would get the solution after the event to be predicted happened.

It is possible to use various approximations, for example degeneration of the dependency of the predicted variable on other events that is then extrapolated to the future. Finding such approximation can be also difficult. This approach generally means creating the model of the predicted event.

Neural networks can be used for forex signal prediction with various levels of success. The advantage to forex systems includes automatic learning of dependencies only from measured data without any need to add further information (such as type of dependency like with the regression). The forex neural network is trained from the historical data with the hope that it will discover hidden dependencies and that it will be able to use them for predicting into future. In other words, neural network is not represented by an explicitly given model. It is more a black box then a forex robot.

It is possible to predict various types of data, however in the rest of this text we will focus on predicting the time series. Time series shows the development of a value in time. Of course, the value can be influenced by other factors than just time. Time series represents discrete history of a value and from a continuous function it can be obtained using sampling.

The prediction type can be classified according to various criteria. Basic criteria are

forex signal data that has a teaching prediction and fx signals with a value or trend.

When we want to get exact value (or more values) of a variable in future, then we are predicting value. The other possibility is to predict trend of a variable, i.e., whether the value will go up or down without considering size of the change - then we are predicting trend. When predicting trend we are in fact classifying into two (or three) classes - up and down (or no significant change). Prediction of a close value is generally easier than predicting a trend. Besides trend we may also want to predict other parts of the trend, such as moving average change etc.

Data for currency exchange predictions.

For time series predicting we usually have values of a variable in equidistant intervals - then we can try to predict the development of the value based on historical values and time only. In this case the historical time series should be long enough and dense enough.

We can have additional information to time series, such as derivation. This information can be then used for more exact prediction. Important information can be added using so called intervention variables (intervention indicators), which represent information about time series or information about the period into which we predict. For example when predicting energy consumption then knowing whether we predict for Monday or Saturday can improve the prediction dramatically - this information does not follow from time series explicitly and must be added. It is usually very helpful to use the values of intervention indicator when creating a model that will be used for prediction.

We can also have information about other related variables, preferably also in time series. From the history of related variables we can reason about other variables. The relation can be expressed in various ways. An example is a static (or slowly changing) sum of two variables. It does not have to be expressed explicitly - for example changes of prices of currency pairs of in one sector are dependent, but can be hard to express in a computational way. This kind of information is selected in a hope that it will cohere with the predicted value, but we do not have to be sure about it. The field of data mining can help with selecting appropriate information and its interpretation.

The advantage of the usage of neural networks for forex signal prediction is that they are able to learn from examples only and that after their learning is finished, they are able to catch hidden and strongly non-linear dependencies, even when there is a significant noise in the training set.

The disadvantage is that NNs can learn the dependency valid in a certain period only. The error of prediction cannot be generally estimated.

We have time series, i.e., a variable x changing in time xt (t=1,2,...) and we would like to predict the value of x in time t+h.

The prediction of time series using forex neural network consists of teaching the net the history of the variable in a selected limited time and applying the taught information to the future. Data from past are provided to the inputs of neural network and we expect data from future from the outputs of the network (see the figure 2).

As we can see, the teaching with teacher is involved. For more exact prediction, additional information can be added for teaching and prediction, for example in the form of intervention's variables (intervention indicators). However, more currency exchange information does not always mean better forex signal prediction; sometimes it can make the process of forex teaching and predicting worse. It is always necessary to select really relevant information, if it is available.

Various types of forex neural networks can be used for prediction, such as back-propagation, ART, Marks network and others. In the rest of this text we will focus on back-propagation.

We have time series, i.e., a variable x changing in time xt (t=1,2,...) and we would like to predict the value of x in time t+h.

The prediction of time series using forex neural network consists of teaching the net the history of the variable in a selected limited time and applying the taught information to the future. Data from past are provided to the inputs of neural network and we expect data from future from the outputs of the network (see the figure 2).

As we can see, the teaching with teacher is involved. For more exact prediction, additional information can be added for teaching and prediction, for example in the form of intervention's variables (intervention indicators) . However, more information does not always mean better prediction; sometimes it can make the process of teaching and predicting worse. It is always necessary to select really relevant information, if it is available.

Various types of neural networks can be used for prediction, such as back-propagation, ART, Marks network and others. In the rest of this text we will focus on back-propagation.Available data are often divided into three set: learning set, validating set and testing set. These sets can overlap and do not have to be continuous. The learning set is a sequence that is shown to the forex neural network during the learning phase. The network is adapted to it to achieve required outputs (in other words, weights in the currency exchange network are changed based on this set). The difference is measured using the validating set and this difference is used to validate whether the learning of the network can be finished. The last set, testing set, is then used to test whether the network is able to work on the data created in the previous process.

To summarize, the learning set is used for creating a model, validation set is used for verifying the model, and the testing set is used for testing of the usability of the model.

Data reprocessing is important as well. For example, it can be useful to remove trend and other components (such as seasonal trends) - of course only if we are able to detect such components. The overview about time series decomposition can be found in references

Especially for forex neural networks that can have outputs only in a certain interval it is important to realize that it is not possible to predict values outside of this interval. Data normalization is then required for the network to be able to get meaningful outputs.

For the illustration of the topic of predicting with neural networks Java applets is available. These forex system applets illustrate the creation of training set and show the result of prediction of the function x=f(t) or of some selected predefined data using neural network of back-propagation type.

The forex software applet enables experimenting with the prediction of time series using back-propagation neural network. A function or data including noise can be used a a base for time series to be learned and predicted. It is possible to set parameters of training set creation and the neural network parameters. The result, i.e., the predicted value, is then compared to the expected value in the future.

Neural networks are suitable for predicting time series mainly because of learning only from examples, without any need to add additional information that can bring more confusion than prediction effect. Neural networks are able to generalize and are resistant to noise. On the other hand, it is generally not possible to determine exactly what a neural network learned and it is also hard to estimate possible prediction error.

Neural networks are being successfully used for predicting time series. Check out a good forex system review website to learn about the new automated forex trading systems using Neural networks. If you check our Forex software review website you can log into a Live Forex managed account and judge for yourself if this new forex robots will help you in your foreign currency exchange trading.