source: orange/docs/widgets/rst/evaluate/liftcurve.rst @ 11050:e3c4699ca155

Revision 11050:e3c4699ca155, 2.3 KB checked in by Miha Stajdohar <miha.stajdohar@…>, 16 months ago (diff)

Widget docs From HTML to Sphinx.

Line 
1.. _Lift Curve:
2
3Lift Curve
4==========
5
6.. image:: ../icons/LiftCurve.png
7
8Shows the lift curves and analyzes them.
9
10Signals
11-------
12
13Inputs:
14   - Evaluation Results (orngTest.ExperimentResults)
15      Results of classifiers' tests on data
16
17
18Outputs:
19   - None
20
21Description
22-----------
23
24Lift curves show the relation between the number of instances which were predicted positive and those of them that are indeed positive. This type of curve is often used in segmenting the population, e.g., plotting the number of responding customers against the number of all customers contacted. Given the costs of false positives and false negatives, it can also determine the optimal classifier and threshold.
25
26.. image:: images/LiftCurve.png
27
28Option :obj:`Target class` chooses the positive class. In case there are more than two classes, the widget considers all other classes as a single, negative class.
29
30If the test results contain more than one classifier, the user can choose which curves she or he wants to see plotted. :obj:`Show convex lift hull` plots a convex hull over lift curves for all classifiers. The curve thus shows the optimal classifier (or combination thereof) for each desired TP/P rate. The diagonal line represents the behaviour of a random classifier.
31
32The user can specify the cost of false positives and false negatives, and the prior target class probability. :obj:`Compute from Data` sets it to the proportion of examples of this class in the data. The black line in the graph, which corresponds to the right-hand axis, gives the total cost for each P ration for the optimal classifier among those selected in the list box on the left. The minimum is labelled by the optimal classifier at that point and the related cost.
33
34The widget allows setting costs from 1 to 1000. The units are not important, as are not the magnitudes. What matters is the relation between the two costs, so setting them to 100 and 200 will give the same result as 400 and 800.
35
36Example
37-------
38
39At the moment, the only widget which give the right type of the signal needed by the Lift Curve is `Test Learners <TestLearners.htm>`_. The Lift Curve will hence always follow Test Learners and, since it has no outputs, no other widgets follow it. Here is a typical example.
40
41.. image:: images/ROCLiftCalibration-Schema.png
Note: See TracBrowser for help on using the repository browser.