Understanding of precision and recall: a. Given the following table of relevant
ID: 653707 • Letter: U
Question
Understanding of precision and recall:
a. Given the following table of relevant documents found in the top 100 documents returned (e.g., 9 rel docs in the first 10, 17 rel docs in the first 20) calculate the Precision/recall graph for the search. Roughly draw a precision recall graph with precision on the Y axis and recall on the X-axis. Assume that all relevant documents are returned by the query. Compare it to the best possible precision/recall graph (draw optimum graph)
1-10
11-20
21-30
31-40
41-50
51-60
61-70
71-80
81-90
91-100
8
9
7
6
4
2
4
2
0
1
1-10
11-20
21-30
31-40
41-50
51-60
61-70
71-80
81-90
91-100
8
9
7
6
4
2
4
2
0
1
Explanation / Answer
Generating a PR curve is {analogous|is comparable} to generating an mythical creature curve. To draw such plots you would like a full ranking of the take a look at set. to form this ranking, you would like a classifier that outputs a call valuerather than a binary answer. the choice worth may be a live of confidence in an exceedingly prediction that we are able to use to rank all take a look at instances. As associate degree example, the choice values of provision regression and SVM ar a likelihood and a (signed) distance to the separating hyperplane, severally.
If you get rid of call values you outline a group of thresholds on aforementioned call values. These thresholds ar totally different settings of a classifier: e.g. you'll management the amount of political orientation. For provision regression, the default threshold would be f(x)=0.5 however you'll check the whole vary of(0,1). Typically, the thresholds ar chosen to be the distinctive call values your model yielded for the take a look at set.
At every alternative of threshold, your model yields totally different predictions (e.g. totally different variety of positive and negative predictions). As such, you get a group of tuples with totally different exactness and recall at each threshold, e.g. a group of tuples (Ti,Pi,Ri). The PR curve is drawn supported the (Pi,Ri) pairs.
Related Questions
Navigate
Integrity-first tutoring: explanations and feedback only — we do not complete graded work. Learn more.