As machine learning techniques become more and more important to human decision-making, the tools to train models are becoming increasingly simple to use at ever higher levels of abstraction. But, while an ever-increasing number of commercial products claim to "take the guesswork out of predictive modeling," practitioners know that machine learning is more complicated than simply picking the best tool or the right algorithm. In practice, training a model is rarely straightforward, requiring several iterations through feature selection and engineering, model fitting, and hyperparameter tuning. While some of this workflow can be automated with grid search, standardized APIs, and GUI-based applications, human steering is almost always more effective than exhaustive search.

Yellowbrick is a new Python library that supports an iterative and interactive human steering process by extending the Scikit-Learn API with visual analysis and diagnostic tools. In particular, Yellowbrick exposes a visual transformer object that can incorporate visualizations of the model selection process into Scikit-Learn pipelines. The Yellowbrick API also wraps Matplotlib to create publication-ready figures and interactive data explorations while still allowing developers fine-grain control of figures. For users, Yellowbrick can help evaluate the performance, stability, and predictive value of machine learning models, and assist in diagnosing problems throughout the machine learning workflow.

In this talk, we'll see not only what you can do with Yellowbrick, but also how it works under the hood -- Yellowbrick is not only a fully open source library, but a welcoming and inclusive project that's always seeking new contributors! We'll explore how Yellowbrick extends the Scikit-Learn and Matplotlib APIs with the Visualizer object, and walk through the process of developing and contributing custom Visualizers of our own.