10/03/2026
Data Science Series | Can a Classifier Know When to Say โI Donโt Knowโ?
In this Data Science Series lesson, I show how to design a small classifier that can recognise uncertainty instead of making a confident guess every time.
Using harmless classroom objects, I define the label, measurable features, honest train-test split, simple baseline, two useful error counts, and an abstain rule. I also explain why every photo of the same physical object must stay in one split and why accuracy alone can hide important mistakes.
Pause the video and complete Question 12.10 by drawing six boxes: label, features, split unit, baseline, error counts, and abstain rule. Then test whether your plan can face a genuinely new object without data leakage.
๐ Buy Data Science and Statistics for the Kids โ Combined Edition: https://www.amazon.com/dp/B0HJRW1ZF1
๐ MathClubForKids: https://mathclubforkids.com/
โถ Brainy Dose with Dr Biplab Pal: https://www.youtube.com/
โถ MathClubForKids YouTube: https://www.youtube.com/channel/UCdDlhOP7JdX3cwHSr8Mbbxw