When developing prediction and classification models, depending on the type of response or target variable, we come across two potential type of problems: the target variable is of categorical type (this is a classification type of problem) or the target variable is of a numeric type (this is a regression type of problem). It has been observed that about 70% of the data belongs to problems arising from classification categories and the remaining 30% are regression problems (here is the reference: https://www.topcoder.com/role-of-statistics-in-data-science/). In this chapter, we will provide steps for applying deep learning neural networks for classification problems. The steps are illustrated using the fetal cardiotocograms, or CTGs.

In this chapter, we will cover the following topics:

• A brief understanding of the fetal cardiotocogram (or CTG) dataset
• Steps for data preparation, including normalization, data partitioning, and one-hot encoding
• Creating and fitting a deep neural network model for classification problems
• Evaluating classification model performance and making predictions using the model
• Fine-tuning the model for performance optimization and best practices