In this chapter, the reader will be presented with a very didactic but interesting application that neural networks are suitable for: disease diagnosis. We've discovered so far that neural networks can be very well applied to classification problems, where one wants to automatically assign some record to a certain category. This chapter digs deeper into this by presenting the basics on how to design a classification algorithm using neural networks. The topics covered in this chapter are as follows:
- Foundations of classification problems
- Logistic regression
- Multiple classes vs. binary classes
- Confusion matrix
- Sensibility and specificity
- Neural networks for classification
- Adaptations in Java code
- Disease diagnosis using neural networks
- Diagnosis for cancer
- Diagnosis for diabetes
One thing that neural networks are really good at is classifying records. A very simple perceptron network draws a decision boundary defining whether a data point belongs to a particular region or to another region, where a region denotes a class. Let's take a look at an x–y scatter chart:
The dashed lines explicitly separate the points into classes. These points represent data records that originally had the corresponding class labels. This implies that their classes were already known; therefore, this classification tasks falls into the supervised learning category.
A classification algorithm seeks to find the boundaries between classes in the data hyperspace. Once the classification boundaries are defined, a new data point, with an unknown class, receives a class label according to the boundaries defined by the classification algorithm. The following figure shows an example of how a new record is classified:
According to the current class configuration, the new record's class is Class 3.