Microsoft Azure AI Fundamentals (AI-900)Describe fundamental principles of machine learning on AzureMedium
A medical research team is using a machine learning model to identify rare disease outbreaks. The dataset is highly imbalanced, with very few positive cases (outbreaks) compared to negative cases (no outbreak). The team wants to ensure the model does not miss any potential outbreaks, even if it means having some false alarms. Which type of error should the team prioritize minimizing?
- ATrue Negatives
- BFalse Negatives
- CFalse Positives
- DTrue Positives
Show answer & explanationAnswer & explanation
Correct answer: B. False Negatives
In this scenario, missing a potential outbreak (a False Negative) has severe consequences. The team prioritizes not missing any outbreaks ('does not miss any potential outbreaks'), indicating that minimizing False Negatives is crucial, even at the cost of some False Positives ('even if it means having some false alarms').
Why the other options are wrong
- A. True Negatives are correctly identified non-outbreaks; minimizing these would be counterproductive.
- C. False Positives are 'false alarms'. The question implies these are acceptable to some extent.
- D. True Positives are correctly identified outbreaks; maximizing these is desired, but minimizing is incorrect.
False Negative (Type II Error)
A False Negative occurs when a machine learning model incorrectly predicts a negative outcome when the actual outcome is positive. It's a 'missed opportunity' or a 'missed detection'.
- Model predicts 'No', but the truth is 'Yes'.
- Can have severe consequences depending on the application (e.g., missing a disease, missing fraud).
- Often minimized by tuning models for higher recall.
Memory trick: A 'False Negative' is a 'No' when it should have been a 'Yes' – a truly missed opportunity.