Episode 6 - Data Science Mini-Series with Dexter Energy
Insider's Guide to Energy
00:00
The Importance of Labeled Data in Machine Learning
In a typical supervised machine learning test, you need labels. So with the icing example, I could see how that could change or be a risk to someone. If we have a client who is ramping up an offshore wind park, and there is no historic data, you're not able to model it statistically because you don't have any historic data. And then comes the challenging part of when do I switch from a physical model to a statistical model? Do I do that at 100 days of data or 200 days ofdata or 50 days of data? There really matters how much historic data do I want to use. Ultimately, for data scientists, it's about having as much data as possible
Transcript
Play full episode
