by Rebecca Tessier
This talk will give a brief overview of validation & selection techniques for predictive models and common occurrences of overfitting when building models in python. We'll walk through some strategies to mitigate overfitting and build better models.
Overfitting is a common problem for anyone who builds statistical models in python and it can be challenging to 1) identify that a model is overfitting and 2) to figure out how to maintain good model performance while also maintaining that performance over time.
We'll start by walking through some examples of detecting an overfitting model and then I'll give a brief overview of some of my favourite techniques for dealing with overfitting including using regularization, ensemble methods & cross validation, and techniques for feature generation and selection. Hopefully this talk will give a quick and informative cheat sheet to people starting out or trying to level up their data science skills, so they can have a more critical eye when they're building their next model in python."
About the Author