Avoiding common machine learning pitfalls

The bigger pictureMachine learning has transitioned from a niche pursuit to one with mass appeal. Thanks to the accessibility of modern machine learning tools, it is now very easy to get started in machine learning, yet this ease of use masks the underlying complexities of doing machine learning. This, coupled with a relatively inexperienced community of practitioners, has led to flawed practices, which are reflected in issues such as poor reproducibility within machine-learning-based studies.This tutorial aims to address this problem by educating practitioners about the many things that can go wrong when applying machine learning and providing guidance on how to avoid these pitfalls. However, this is just part of the longer-term process that is needed to improve practice, as machine learning will only meet its ambitions if it is able to become a robust and trusted applied discipline. Other factors that have a role to play in this include better tools, standardization, and regulation.SummaryMistakes in machine learning practice are commonplace and can result in loss of confidence in the findings and products of machine learning. This tutorial outlines common mistakes that occur when using machine learning and what can be done to avoid them. While it should be accessible to anyone with a basic understanding of machine learning techniques, it focuses on issues that are of particular concern within academic research, such as the need to make rigorous comparisons and reach valid conclusions. It covers five stages of the machine learning process: what to do before model building, how to reliably build models, how to robustly evaluate models, how to compare models fairly, and how to report results.