Pipelines

During your data science journey, you will have realized that there are various processes to go through before you get any final outcomes, including data cleaning, data transformation, modeling, and so on. You will have found that when we perform these activities separately, we have to write individual steps to carry them out. When we have to do these steps for a new dataset, even if it is similar to another one that we have worked on before, we have to repeat the steps again. As you will have noticed, these steps are tedious and repetitive.

Pipeline is a utility from the scikit-learn library that automates many of these tasks. Using this utility, we can stack multiple processes together for use on multiple datasets. This helps us with a lot of the manual steps involved in the data science journey.

Have a look at Figure 16.1; we have used pipeline processes on multiple different datasets. However, thanks to the pipeline processes we are provided with, we can be sure...