Structured data involving numeric response variables is classified as a regression problem. For example, the price of a house in a city may depend on variables such as the age of the house, the crime rate in the city, the number of rooms, and the property tax rate. Although statistical methods, such as multiple linear regression and elastic net regression, can also be useful for these situations, deep learning networks offer certain advantages. One of the main advantages of using neural networks in general is that they can capture non-linearity. Unlike statistical methods that require certain assumptions to be met before we can use them, neural network-based models are more flexible to use and do not require many assumptions to be fulfilled.

Many applications involving regression problems also call for identifying variables or features that have a significant impact on the response variable. However with deep learning networks, such feature engineering is inbuilt, and it doesn't call for any extra effort in extracting important features. One thing to note regarding deep learning networks is that the larger the dataset being used, the more effective the resulting prediction model will be. In Chapter 3, Deep Neural Networks for Regression, we provide an illustration of a deep learning regression model using R.