ABSTRACT
This chapter covers two of the most recognized neural net structures. For supervised classification, the focus is on feed-forward neural networks trained with backpropagation. For unsupervised clustering, the discussion centers on Kohonen self-organizing maps. Section 8.1 introduces basic neural network concepts and terminology. Section 8.2 provides a conceptual overview of supervised and unsupervised neural networks. Section 8.3 looks at techniques that have been developed for neural network explanation. Section 8.4 offers a list of general strengths and weaknesses found with all neural networks. Section 8.5 presents detailed examples of how backpropagation and self-organizing neural networks are trained. Section 8.6 features scripts that employ two neural network packages available with R to create and test supervised neural net models. Section 8.7 is all about unsupervised neural net clustering with R. The focus of the final section is on stock market prices and time-series data. Several end of chapter exercises provide practice building neural network models.
