Bài kiểm tra có liên kết hướng dẫn · Trung bình Cấp độ

Bài kiểm tra biểu đồ nâng và tăng

Kiểm tra mức tăng tích lũy, đường cơ sở ngẫu nhiên, mức tăng phân số cao nhất, số học tỷ lệ phản hồi, lựa chọn công suất, giới hạn hiệu chuẩn và cảnh báo nguyên nhân.

Đường dẫn hướng dẫn liên quanBiểu đồ nâng và tăng
Câu hỏi 1 của 8

What does cumulative gain report at a selected fraction of a scored population?

Xem trước mọi câu hỏi8 câu hỏi

Câu trả lời và giải thích xuất hiện khi bạn làm bài kiểm tra.

  1. What does cumulative gain report at a selected fraction of a scored population?
    • The causal effect of contacting everyone in the fraction
    • The average numerical probability assigned to that fraction
    • The fraction of model training records with no label
    • The share of all actual positives found within that top-ranked fraction
  2. In the guide's invented 1,000-person set, the top 200 contain 50 of 100 positives. What is gain at 20%?
    • 20%, the fraction of people selected
    • 25%, the response rate within the selected group
    • 2.5, the lift multiple
    • 50%, the fraction of all positives captured
  3. At a 20% population fraction, what gain is expected from random selection on average?
    • 20%, the diagonal baseline at that fraction
    • 10%, the overall positive rate
    • 50%, the model's observed gain
    • 100%, because all positives can be found eventually
  4. At 20% capacity, how much lift does 50% gain represent?
    • 0.4, dividing 20% by 50%
    • 0.5, the model's gain itself
    • 2.5, dividing 50% gain by the 20% random baseline
    • 0.7, adding gain and capacity
  5. What positive rate does the selected top-200 group have in the constructed example?
    • 10%, the overall base rate
    • 25%, because 50 of 200 selected cases are positive
    • 50%, because half of all positives were captured
    • 2.5%, converting lift to a percentage
  6. Why should two ranking models be compared at the team's real review capacity?
    • Review capacity automatically establishes probability calibration
    • A model's lift is identical at every population fraction
    • Gain and lift can differ by the fraction selected, while the team can act on only that many cases
    • The highest-scored person alone determines total workload
  7. A chart shows high lift among people contacted. What does that fail to prove?
    • That positives were concentrated in the selected score band
    • That contact itself caused those people to respond
    • That an outcome label was available for evaluation
    • That the score sorted some cases ahead of others
  8. Which evidence should accompany a lift multiple before an outreach decision?
    • Counts, base rate, selected fraction, costs and relevant group outcomes
    • Only one chart image without its cutoff
    • Only the model's training runtime
    • An assumption that high lift guarantees profit