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

Đố vui rừng biệt lập

Kết nối độ dài đường dẫn cây ngẫu nhiên với thứ hạng bất thường và phân biệt ngưỡng ô nhiễm với tỷ lệ bất thường trong thế giới thực.

Đường dẫn hướng dẫn liên quanRừng Cách Ly
Câu hỏi 1 của 8

What tends to happen to an unusual point's average isolation-tree path length?

Xem trước mọi câu hỏi8 câu hỏi

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

  1. What tends to happen to an unusual point's average isolation-tree path length?
    • It is shorter because fewer random splits isolate it
    • It is longer because every split keeps it with common points
    • It equals the number of features exactly
    • It is set by contamination directly
  2. What does the contamination setting commonly control?
    • The verified prevalence of anomalies in production
    • The threshold used to label a chosen fraction as outliers
    • The number of input features
    • The random split direction for every tree
  3. Which score interpretation matches the forest's core intuition?
    • Longer paths always mean higher anomaly risk
    • A score is a calibrated fraud probability
    • Shorter paths indicate stronger isolation-based anomaly evidence
    • The score identifies the causal reason for an event
  4. A rare but legitimate customer segment receives high anomaly scores. What issue does this illustrate?
    • Path lengths are always causal
    • Contamination guarantees correct labels
    • Random trees require class labels
    • Rarity is not equivalent to fraud or invalidity
  5. Why verify score direction in the chosen library?
    • All libraries define negative scores as normal
    • Some decision functions invert or shift the raw anomaly score
    • The tree ensemble omits path lengths
    • Contamination reverses the feature matrix
  6. If contamination increases from 5% to 10%, what change is most directly expected?
    • Every path becomes shorter
    • The true anomaly rate is proven to double
    • More points may be labeled anomalous by the threshold
    • The model changes into supervised fraud classification
  7. How should an alerting team evaluate an Isolation Forest threshold when labels are available?
    • Choose the threshold that labels exactly half the data
    • Assume every short path is fraud
    • Compare only the training score average
    • Measure precision and recall tradeoffs on representative labeled cases
  8. What can an Isolation Forest score not provide by itself?
    • A causal explanation or calibrated event probability
    • A ranking of isolation behavior
    • An ensemble-based anomaly score
    • A comparison across observations