Subject
Computer Science
Topic
Artificial Intelligence
A feature computed after the prediction time is included during training. What problem does this create?
A feature computed after the prediction time is included during training. What problem does this create?
a)
Feature scalingb)
Data leakagec)
Regularizationd)
Under-sampling onlyExplanation
Information unavailable at real prediction time leaks the answer into training and produces unrealistically optimistic evaluation.Related questions
A classifier has TP=40, FP=10, FN=5, TN=45. What is its accuracy rounded to two decimals?A classifier has TP=30, FP=10, FN=20, TN=40. What is its precision rounded to two decimals?A classifier has TP=36, FP=4, FN=9, TN=51. What is its recall rounded to two decimals?A classifier has TP=24, FP=6, FN=6, TN=64. What is its f1 rounded to two decimals?In the figure, a binary node contains class counts 6 and 2. What is its entropy in bits, rounded to three decimals?In the figure, a split changes parent counts 6,4 into left 4,1 and right 2,3. What is information gain rounded to three decimals?A condition has prior 0.1, test sensitivity 0.8, and false-positive rate 0.2. What is P(condition given positive), rounded to three decimals?A rare class must appear proportionally in both training and test data. Which split is appropriate?