Avoid These Easily Missed Mistakes in Machine Workflow – Part 1 | by Thomas A Dorfer | January, 2025
Misusing identifiers, misclassifying data, and ignoring unusual attribute values
One of the most exciting things about being part of the machine learning community for as long as I have is the opportunity to learn something new every now and then. That something new can be a tool or a new method (given the rapid development in the field of machine learning, there is never a shortage of that), but it can also be the discovery of faulty processes in our work that we have not seen before. of.
Some of these may be vague and hard to see at first. If these error processes enter the development of your model, there is a good chance that it will damage its predictive power and thus its reliability, and, ultimately, its performance.
In this article, which is the start of a series examining common pitfalls in machine learning, we will focus on three data handling errors that can occur both during the pre-processing phase but also during the modeling phase:
- Using Numeric Identifiers as Features
- Random Classification Instead of Group Classification
- Including Attribute Values and Incomplete Observations
