Topic 20

Question 1

Confusion Matrix

Answer 1

A table that counts the number of predictions of each class divided up by the actual class labels):

Question 2

Precision

Answer 2

number of true positives /
number of true positives + number of false positives (accuracy of the model predicting yes)

Question 3

Recall / Sensitivity

Answer 3

number of true positives /
number of true positives + number of false negatives (proportion of yes that were predicted correctly)

Question 4

F1 Score

Answer 4

2 ∗ precision × recall /
precision + recall

Question 5

Cohen’s Kappa

Answer 5

k = 1 - 1-p0 / 1-pe where po is the proportion of times our model is correct and pe is the propotion of times a random model is correct

Question 6

Specificity

Answer 6

Number of true negatives / number of true negatives + number of false positives (accuracy of the model predicting no)

Question 7

Accuracy

Answer 7

number of true positives + number of true negatives /
number of true positives + number of true negatives+number of false positives + number of false negatives (combination of precision, recall and specificity)

Question 8

ROC Curve

Answer 8

A plot considering every possible decision boundary and comparing the true positive and false positive rate

Question 9

Area Under the Curve (AUC) for ROC Curve

Answer 9

The proportion of 1x1 unit square that falls below the ROC curve (higher is better)