KwickClips Artificial Intelligence · 39 sec · free
Should you trust 99% on training data?
Short answer. No. Test only on unseen data.
What is a common split?
80% train, 20% test.
Split the data
| Training set: model learns |
| Testing set: kept hidden |
| Often 80% and 20% |
Remember
| Test only on unseen data |
Your model scores ninety nine percent. Should you trust it? Only if tested fairly. The model learns from the training set. The testing set stays hidden. A common split is eighty and twenty. Test only on data the model has not seen.
This clip is from the full lesson: Data Pre-processing and Data Interpretation — 7 minutes, with the tables, the quick answers and the whole lesson in text.
Useful for: CBSE Class 9 Artificial Intelligence (417), CBSE Class 11 Artificial Intelligence (843), CBSE Class 11 Artificial Intelligence (843)
More KwickClips from this lesson
What do you do with a duplicate row?38 sec
Can a model read the word Delhi?42 sec
Does ice cream cause heat stroke?37 secVoice-over is AI-generated; the script is written and checked by Kajal Ma'am. Confirm anything you plan around against your official board document. We never ask for a password or an OTP.