Quiz połączony z przewodnikiem · Średni Poziom

Quiz o lokalnych czynnikach odstających

Zinterpretuj LOF jako współczynnik gęstości lokalnej i odróżnij go od odległości globalnej, w tym wyborów w skali sąsiada i trybu nowości.

Powiązane ścieżki prowadząceLokalny czynnik odstający
Pytanie 1 z 8

What does an LOF score substantially above one suggest?

Podgląd każdego pytania8 pytania

Odpowiedzi i wyjaśnienia pojawiają się w miarę rozwiązywania quizu.

  1. What does an LOF score substantially above one suggest?
    • The point is less dense than its local neighbors
    • The point is equally dense as its local neighbors
    • It is far from the global center even if its neighborhood is dense
    • The score is a calibrated anomaly probability
  2. Why can a far-away point receive an ordinary LOF score?
    • LOF ignores all neighbor information
    • It may belong to a separate dense subgroup with similar local density
    • Distance is always set to zero outside the main cloud
    • LOF measures only the global mean
  3. What does increasing the neighbor count generally change?
    • The number of features used in the density ratio
    • The sign convention of the returned score
    • The scale over which local density is compared
    • The global clustering method used before LOF
  4. Why does feature scaling matter for LOF?
    • It changes local density into a Gaussian parameter
    • It guarantees every feature contributes equally without preprocessing
    • It changes neighbor count into a class label
    • Nearest-neighbor distances depend on feature units
  5. How does novelty detection differ from fitted-sample outlier detection?
    • It requires labels for every training sample
    • It scores new observations against a reference fit
    • It retrains a new density for each query
    • It reports only global cluster means
  6. In scikit-learn novelty mode, what should users avoid according to its documentation?
    • Checking neighbor count
    • Scaling numeric predictors
    • Using novelty scoring methods on the training set
    • Scoring new unseen cases
  7. Which pattern indicates a local rather than global anomaly concept?
    • Distance from the global center is large even when local density is high
    • The observation has a rare category label
    • The global cluster contains few observations
    • Density is low relative to nearby points
  8. Why check the sign convention of an LOF API score?
    • Libraries may invert scores so stronger outliers have lower numeric values
    • The sign determines how many neighbors exist
    • LOF scores are always probabilities
    • The score direction is identical in every implementation