ABSTRACT
This chapter describes a popular machine learning method for epi-forecasting called neural networks. It starts with an introduction to the building blocks of a neural network, and then it describes how a neural network can be trained. The chapter also presents a class of neural network autoregressive models to make predictions for time series data, which combines traditional statistical models and neural networks. It demonstrates the usage of neural network autoregressive models with an application to COVID-19 data. The basic computation unit in a neural network is the node or unit patterned after a neuron in a human brain. A neural network training algorithm finds the parameters with the help of forward propagation and backpropagation. Often neural networks have too many weights and will overfit the data. A solution to overcome this overfitting issue is to update the training algorithm and encourage the network to keep the weights small.
