Decision trees build tree structures to generate regression or classification models. You can think of it as a collection of chained if, and else if statements that will culminate in predictions. These models are very flexible:
- Categorical and numerical input/output is welcomed
- Classifications and regressions can be made using tree-based models
- Trees can grow very long (and complicated) or small (and simple)
Although it's possible to design very complicated trees, it is not recommended to do so. Over-complicated trees tend to be a great source of overfitting.
Needless to say, it is very easy to implement tree-based models with R. Even more complex algorithms that rely on trees as basic building blocks can be easily implemented with R (more on that in the Random forests – a collection of trees,). The current section will make a quick tour through tree-based models using R. Strengths and weakness will be put under the spotlight.