←

Training, testing and evaluating models

From Mock A · Paper 1: Written examination
61 mark
Why is test data kept separate from training data?
From Mock A · Paper 1: Written examination
14(a)2 marks
A model is tested on 200 new images of cats and dogs. It classifies 170 of them correctly. Calculate the accuracy of the model as a percentage. Show your working.
From Mock A · Paper 1: Written examination
14(b)2 marks
The same model scored 98% on its training data but 85% on the test data. Explain what this suggests about the model.
From Mock A · Paper 1: Written examination
14(c)2 marks
Suggest two ways the developers could improve the model's performance on new images.
From Mock A · Paper 1: Written examination
199 marks
A hospital plans to use an AI system to help decide which patients arriving at the emergency department should be seen first. The system was trained on records from previous patients. Evaluate whether the hospital should use this system.
From Mock A · Paper 2: Combined internal assessment (project)
3(b)2 marks
Explain how you separated data for training and testing, and why.
From Mock B · Paper 1: Written examination
61 mark
A self-driving car's model was trained mostly on photos taken in daylight. It performs poorly at night. What is the best explanation?
From Mock B · Paper 1: Written examination
15(a)1 mark
A model predicts whether a piece of fruit is ripe. It is tested on new fruit:
Predicted ripePredicted not ripe
Actually ripe428
Actually not ripe644
State how many pieces of fruit were tested.
From Mock B · Paper 1: Written examination
15(b)2 marks
Calculate the accuracy of the model in 15(a). Show your working.
From Mock B · Paper 1: Written examination
15(c)2 marks
A supermarket wants to avoid selling unripe fruit. Explain which type of error in the table matters most to the supermarket.
From Mock B · Paper 2: Combined internal assessment (project)
3(b)2 marks
Explain how you separated data for training and testing, and why.
From Mock C · Paper 1: Written examination
71 mark
What is overfitting?
From Mock C · Paper 1: Written examination
11(b)3 marks
Describe how the developers should test the model in 11(a).
From Mock C · Paper 1: Written examination
14(d)2 marks
State the value returned by `classify_temp(10)` and explain why 10 is an important value to test.
From Mock C · Paper 1: Written examination
15(b)2 marks
The model often classifies crisp packets as recyclable foil. Suggest two ways to improve it.
From Mock C · Paper 2: Combined internal assessment (project)
3(b)2 marks
Explain how you separated data for training and testing, and why.
From Mock D · Paper 1: Written examination
71 mark
What is training data?
From Mock D · Paper 1: Written examination
11(b)2 marks
Explain why the spam filter needs to be retrained regularly.
From Mock D · Paper 1: Written examination
11(c)2 marks
The filter wrongly sends an important job offer to the spam folder. Name this type of error and explain its impact.
From Mock D · Paper 1: Written examination
11(d)3 marks
The developers have 10 000 labelled emails. Describe how they should split the data and explain why.
From Mock D · Paper 2: Combined internal assessment (project)
3(b)2 marks
Explain how you separated data for training and testing, and why.
From Mock E · Paper 1: Written examination
71 mark
A model correctly classifies 45 out of 60 test items. What is its accuracy?
From Mock E · Paper 2: Combined internal assessment (project)
3(b)2 marks
Explain how you separated data for training and testing, and why.
From Mock F · Paper 1: Written examination
71 mark
Why should developers record performance outcomes when testing a model?
From Mock F · Paper 1: Written examination
11(d)2 marks
The model is 90% accurate. Only 10% of players are injured each season. Explain why 90% accuracy may be misleading.
From Mock F · Paper 1: Written examination
209 marks
A secondary school plans to use an AI system to mark students' essays automatically and predict their final exam grades. Evaluate this plan.
From Mock F · Paper 2: Combined internal assessment (project)
3(b)2 marks
Explain how you separated data for training and testing, and why.
Score 0 / 63 · 0 of 27 marked

This paper is written by Plasmo in the style of the LRN GCSE Artificial Intelligence exam. It is not a past paper and is not produced or endorsed by the exam board.

Training, testing and evaluating models Exam Questions: LRN GCSE Artificial Intelligence | Plasmo