Scikit-learn is like the Swiss Army Knife of data science and API consistency is key to its success. Amongst a massive amount of components, it provides an easy to use pipeline object which is very useful in organizing your machine learning pipelines incl. also model selection and grid search. Despite the great variety of sklearn methods, in real-life business applications there is always a need for new building blocks (e.g. to combine multiple models that are trained on different sub-sets of the data into a single pipeline or to add some simple business rules on top of your classifier).
In this presentation, we will show how you can develop your own custom components that are fully compatible with scikit-learn, so you can enrich your sklearn pipelines in the best possible way without losing any of its API consistency. So, after the talk you will be able to develop more creative modelling architectures than you were able before!