BITCOIN FORECASTER WITH NEURAL NETWORKS

Abstract

An artificial neural network is a model from the field of machine learning that was inspired by the structure and function of the brain.Whilst deep learning involves neural networks with multiple layers between the input and output layers.This paper shows how neural networks work, and describes various features.Backpropagation is explained in detail, and the common issues that one can face whilst training a neural network are described.Methods of mitigating these issues are also given.Finally, a neural network is employed to build a daily Bitcoin trading system.The results were inferior to multiple linear regression.The neural network likely suffered from overfitting, but regularisation using dropout helped to mitigate this.