Data Mining MCQs

Multiple binary classifiers are trained and their outputs combined.
Libraries typically implement one of these strategies automatically.

Values near one indicate a well-matched cluster assignment.
Values near or below zero suggest overlap or possible misassignment.

Neighborhood updates encourage nearby grid units to represent similar regions.
This creates topological organization.

The method distinguishes the normal data region from the origin or outside space.
New points beyond the learned boundary are flagged.

Structures such as k-d trees partition feature space for faster queries.
Their effectiveness can decline in very high dimensions.

Each classifier separates one target class from the rest.
The class with the strongest decision score is often selected.

K-medians minimizes an L1-style objective rather than squared error.
Medians are less sensitive to extreme values than means.

Large early neighborhoods organize the whole map coherently.
Smaller later neighborhoods refine local prototypes.

Precision indicates how many alerts are truly anomalous.
Low precision creates operational cost and alert fatigue.

Encoding red, blue, and green as 1, 2, and 3 creates artificial spacing.
Appropriate one-hot encoding or mixed-type distances should be used.

Methods such as Platt scaling map decision scores to probabilities.
Calibration should use separate or cross-validated data.

Absolute deviations and median centers reduce the influence of extremes.
This can produce more robust cluster locations.