Anomaly Detection Flashcards
1
Q
DTF never results in distance strictly greater than Euclidean distance
A
True
2
Q
DTF cannot be applied to sequences of diff length
A
False
3
Q
DTF can only be applied to single-variate (one-dimensional/ one feature) sequence
A
False
4
Q
DTF normalization is useful
A
True
5
Q
What does a low local reachability density mean? (lrd)
A
It means large average distance
6
Q
LOF(q) < 1 means what?
A
Inlier, higher density
7
Q
LOF(q) > 1 means what?
A
Outlier, lower density
8
Q
Advantages of NN?
A
- used in unsupervised setting
- no assumptions about data distribution
- intuitively appealing, uses distances
9
Q
Disadvantages of NN?
A
- computationally expensive when testing
- requires distances, so all disadvantages of distances apply
10
Q
Advantages of PCA?
A
- Useful for modeling feature interaction
- Computationally efficient
11
Q
Disadvantages of PCA?
A
- Based on assumption that normal/ anomaly are distinguishable in the reduced space
- Context not taken into account
- PCA sensitive to outliers
12
Q
What are the three types of anomalies?
A
- Point (point x is strange)
- Contextual (point x strange given set S)
- Collective (set S is strange)
13
Q
Outliers have no effect on PCA?
A
False
14
Q
PCA assumes relationship between variables is linear?
A
True
15
Q
LOF uses reachability distance instead of actual distance to lower effect of outliers?
A
False
