Bài kiểm tra có liên kết hướng dẫn · Trung bình Cấp độ
Tỷ lệ cơ bản và kết quả dương tính giả trong bài kiểm tra phát hiện AI
Hoạt động thông qua mức độ phổ biến, độ nhạy, tỷ lệ dương tính giả, giá trị dự đoán, ngưỡng và lý do tại sao cờ của trình phát hiện yêu cầu xem xét theo ngữ cảnh.
Câu hỏi 1 của 8
A condition is rare in a population. Why can a detector with a modest false-positive rate produce many false alarms?
Xem trước mọi câu hỏi8 câu hỏi
Câu trả lời và giải thích xuất hiện khi bạn làm bài kiểm tra.
- A condition is rare in a population. Why can a detector with a modest false-positive rate produce many false alarms?
- There are many more unaffected cases to generate false positives
- Sensitivity becomes zero
- Every negative case is counted as positive
- Prevalence cancels from the calculation
- In the guide's example, 1 percent prevalence among 1,000 items means how many actual positive items?
- 50
- 10
- 90
- 100
- What does sensitivity measure?
- The share of all flags that are correct
- The condition's prevalence
- The share of actual positive cases that the detector flags
- The share of negative cases incorrectly flagged
- Which question does positive predictive value answer?
- Among positive cases, what share are flagged?
- How common is the condition before testing?
- How many thresholds exist?
- Among flagged cases, what share are truly positive?
- Why should a benchmark's evaluation population be compared with the deployment population?
- Different prevalence can change the meaning of a positive flag
- Benchmarks never use labels
- It makes sensitivity unnecessary
- The detector will otherwise refuse to run
- What commonly happens when a detector threshold is lowered?
- It eliminates every false positive
- It may catch more positives while producing more false positives
- It leaves sensitivity unchanged by definition
- It changes prevalence
- For a consequential decision, how should a detector flag be used?
- As conclusive proof of the condition
- As a replacement for an appeal process
- As a prompt for additional evidence and review
- As a reason to ignore context
- Why report an evaluation threshold with error rates?
- Because thresholds determine prevalence
- Because sensitivity is unrelated to thresholds
- Because a single accuracy score contains all details
- Because the threshold defines the operating tradeoff between misses and false alarms