Abstract
Artificial neural Networks (ANNs) are a powerful technique for multivariate dependence analysis. Originally inspired by neuroscience, ANNs are becoming an increasingly attractive analytic tool for applications in the area of economics and finance due to the flexible solutions they offer. The purpose of this article is to present such important applications with an emphasis on recent research trends. The contributions are grouped as follows: ANNs (1) for prediction, (2) for classification and (3) for modelling. The chapter concludes with the future trends in the ANNs research in economics and finance.
TopMain Focus Of The Article
The purpose of this section is to survey important applications of ANNs in economics and finance. Emphasis has been laid on recent research trends in contributions in which ANNs are the main empirical method or are combined with alternative research tools.
Key Terms in this Chapter
Multicollinearity: This concept explains the situation in regression analysis when two or more of the independent variables are moderately to highly correlated. Consequently, the regression results are biased.
Stationarity: A time series is stationary when its statistical distribution moments do not change over time.
Supervised Learning: It consists in learning from data with a known-in-advance outcome that is predicted based on a set of inputs, referred to as “features”.
Explanatory Capacity of Artificial Neural Networks: It consists in the implementation of methods that are able to determine the relative contribution of each ANN input variable to the ANN output. See Gevrey et al. (2003) for further details.
Deep Learning: A branch of machine learning to whose architectures belong deep ANNs. The term “deep” denotes the application of multiple layers with a complex structure.
Data Snooping: Denotes the use of statistical inference methods for detecting statistically significant patterns without prior theorizing and formulating hypotheses regarding their causal relationship.
Artificial Neural Network Architecture (Topology): This concept involves decision-making concerning the size and number of the ANN layers, as well as the design of the connections among the ANN units.