Deep Learning for Finance

Chapter 8. Deep Learning for Time Series Prediction I

Deep learning is a slightly more complex and more detailed field than machine learning. Machine learning and deep learning both fall under the umbrella of data science. As you will see, deep learning is mostly about neural networks, a highly sophisticated and powerful algorithm that has enjoyed a lot of coverage and hype, and for good reason: it is very powerful and able to catch highly complex nonlinear relationships between different variables.

The aim of this chapter is to explain the functioning of neural networks before using them to predict financial time series in Python, just like you saw in Chapter 7.

A Walk Through Neural Networks

Artificial neural networks (ANNs) have their roots in the study of neurology, where researchers sought to comprehend how the human brain and its intricate network of interconnected neurons functioned. ANNs are designed to produce computational representations of biological neural network behavior.

ANNs have been around since the 1940s, when academics first started looking into ways to build computational models based on the human brain. Logician Walter Pitts and neurophysiologist Warren McCulloch were among the early pioneers in this subject. They published the idea of a computational model based on simplified artificial neurons in a paper.¹

The development of artificial neural networks gained further momentum in the 1950s and 1960s when researchers like Frank Rosenblatt worked on the ...