Introduction to Classification

What is classification?

Moving on from regression models, we’ll now focus on classification methods. As discussed in the introduction to this chapter, these aim to predict class variables that are discrete, categorical, or ordinal. An example could be trying to predict the level of skill of a Dota player based on a set of performance features. In the remaining part of this chapter, we’ll try to predict the variable SkillLevel for players in the Dotalicious dataset using the rest of the performance features in that dataset as input features. We’ll use this as an example in the labs below and show how to build different classification methods to build predictive models. SkillLevel is a discrete numeric variable that is suitable for classification in the previous lab.

Classification algorithms

In the following subsections, we’ll discuss each algorithm. We’ll first outline the main idea and then delve deeper into each method. We’ll also discuss the hyperparameters that each method uses or needs to tune. As described above, these are the parameters that are used as input for the algorithm to learn the model. Hyperparameter tuning refers to the task of picking values for these hyperparameters to achieve the most suitable model to conduct learning on. As done previously, each algorithm will be supplemented with a lab describing how we can use the algorithm with game data.