R Tutorial: Getting started with caret

Want to learn more? Take the full course at https://learn.datacamp.com/courses/supervised-learning-in-r-case-studies at your own pace. More than a video, you'll learn hands-on coding & quickly apply skills to your daily work. --- You just performed some exploratory data analysis and built a simple linear model using base R's lm() function. You were able to see how the fuel efficiency for these cars is distributed and to get an idea of whether you will be able to train accurate models. Now it's time to bring out a more powerful and flexible set of tools for predictive modeling. We are going to use the caret package in this course and the first thing we are going to practice is splitting your data into a training set and a testing set. It is best practice to hold out some of your data for testing in order to get a better estimate of how your models will perform on new data, especially when you use very powerful machine learning techniques. Linear regression doesn't really fall into that category, but we are going to practice this anyway. caret has functions that help you specify training and testing sets. And you can create these so that they balance some characteristic in your dataset. For example, the code here takes an input data set and puts 80% of it into a training dataset and 20% of it into a testing dataset; it chooses the individual cases so that both sets are balanced in aspiration types. Why are we bothering with this? The point of holding data back from the model training process is to have something to test that data on. We want to be able to estimate how well our model will perform on new data, and the best way to do is to use data that was not an input to training the model. Holding out testing data allows you to assess if your model is overfitting. It's also possible to divide your data into three partitions as you build, choose, and assess models, and we'll talk about this later in the course. Once you have a training dataset, you can train