Market Modeling, Forecasting and Risk Analysis with Historical Consistent Neural Networks

摘要

Business management requires precise forecasts in order to enhance the quality of planning throughout the value chain. Furthermore, the uncertainty in forecasting has to be taken into account. Neural networks (NN) offer significant benefits for dealing with the typical challenges associated with forecasting. With their universal approximation properties, NN make it possible to describe non-linear relationships between a large number of factors and multiple time scales Pt In contrast, conventional econometrics (such as ARMA, ARIMA, ARMAX) remain confined to linear systems [8]. A wide range of models is discussed within the class of neural networks. For example, in terms of the data flow, it is possible to draw a distinction between feedforward and (time) recurrent NNs [1]. In this paper we focus on recurrent NN.