In this chapter, we explored classification and the various algorithms that can be used to model a given classification problem. Although classification techniques are very useful, they do not perform too well when the sample data has a large number of dimensions. Also, the features may vary in a nonlinear manner, as we will describe in Chapter 4, Building Neural Networks. We will also explore more about these aspects and the alternate methods of supervised learning in the following chapters. The following are few of the points that we looked at in this chapter:
- We described two broad types of classifications, namely, binary and multiclass classification. We also briefly studied the logistic function and how it can be used to model classification problems through logistic regression.
- We studied and implemented a Bayes classifier, which used a probabilistic model used for modeling classification. We also described how we could use the
clj-mllibrary's Bayes classifier implementation to model a given classification problem. - We also explored the simple k-nearest neighbor algorithm and how we can leverage it using the
clj-mllibrary. - We studied decision trees and the C4.5 algorithm. The
clj-mllibrary provides us with a configurable implementation of a classifier based on the C4.5 algorithm, and we described how this implementation could be used as well.
We will explore artificial neural networks in the following chapter. Interestingly, we can use artificial neural networks to model regression and classification problems, and we will study these aspects of neural networks as well.