metrics Flashcards
1
Q
What is type 1 error
A
false positive rate
2
Q
What is type 2 error
A
false negative rate
3
Q
What is precision
A
TP / ( TP + FP )
4
Q
What is recall
A
TP / ( TP + FN )
5
Q
What is accuracy
A
(TP + TN) / all
6
Q
What is the ROC curve
A
receiver operating characteristic curve
x is FPR = 1 - specificity
y is TPR = sensitivity = recall
7
Q
What is f-score
A
2(precisionrecall) / (precision+recall)
8
Q
What is bias
A
model bias = underfitting
not complex as data
9
Q
What is variance
A
model variance = overfitting
10
Q
What is overfitting
A
model has high variance
model is more complex than data
11
Q
What is underfitting
A
model has bias
model is less complex than data
12
Q
Increase overfitting is an increase in?
A
variance
13
Q
Increase underfitting is an increase in?
A
bias
14
Q
What is regularization for
A
- control model complexity
- help generalization
- add penality to cost function
15
Q
L1 regularization is
A
- L1 norm = sum of absolute beta weights
- feature selection
- add to min error cost function
