Prediction
Predicting is making claims about something that will happen, often based on information from the past and from the current state.
Everyone solves the problem of prediction every day with various degrees of success. For example, weather, harvests, energy consumption, movements of forex (foreign exchange) currency pairs or stock prices, earthquakes, and many other things need to be predicted.
In the technical domain, predictable parameters of a system can often be expressed and evaluated using equations - prediction is then simply the evaluation or solution of such equations. In practice, however, we often face problems where such a description would be too complicated or not possible at all. In addition, solving the problem in this way can be computationally demanding, and sometimes the result would arrive only after the predicted event has already happened.
It is possible to use various approximations, for example regression of the dependence of the predicted variable on other factors, which is then extrapolated into the future. Finding such an approximation can also be difficult. This approach generally means building a model of the predicted phenomenon.
Neural networks can be used for prediction with various levels of success. The advantages of them include automatic learning of dependencies directly from measured data, without the need to add further information (such as the type of dependency, as in regression). The neural network is trained on historical data with the hope that it will discover hidden dependencies and that it will be able to use them for future prediction. In other words, a neural network is not represented by an explicitly defined model. It is more like a black box that is able to learn useful relationships.
It is possible to predict various types of data, however in the rest of this text we will focus on the prediction of time series (see figure 1). A time series shows the development of a value in time. Of course, the value can be influenced by factors other than just time. A time series represents the discrete history of a value, and it can be obtained from a continuous function by sampling.
Figure 1 - Example of time series
ยท