Introduction

In the previous chapters, we learned about building a neural network and the various parameters that need to be tweaked to ensure that the model built generalizes well. Additionally, we learned about how neural networks can be leveraged to perform image analysis using MNIST data.

In this chapter, we will learn how neural networks can be used for prediction on top of the following:

• Structured dataset
  • Categorical output prediction
  • Continuous output prediction

• Text analysis
• Audio analysis

Additionally, we will also be learning about the following:

• Implementing a custom loss function
• Assigning higher weights for certain classes of output over others
• Assigning higher weights for certain rows of a dataset over others
• Leveraging a functional API to integrate multiple sources of data

We will learn about all the preceding by going through the following recipes:

• Predicting a credit default
• Predicting house prices
• Categorizing news articles
• Predicting stock prices
• Classifying common audio

However, you should note that these applications are provided only for you to understand how neural networks can be leveraged to analyze a variety of input data. Advanced ways of analyzing text, audio, and time-series data will be provided in later chapters about the  Convolutional Neural Network and the Recurrent Neural Network.