source: orange/docs/widgets/rst/data/rank.rst @ 11795:7d7ee77fd99b

Revision 11795:7d7ee77fd99b, 2.1 KB checked in by blaz <blaz.zupan@…>, 5 months ago (diff)

Updated documentation for Rank widget.

Line 
1.. _Rank:
2
3Rank
4====
5
6.. image:: ../../../../Orange/OrangeWidgets/Data/icons/Rank.svg
7
8Ranking of attributes in classification or regression data sets.
9
10Signals
11-------
12
13Inputs:
14
15   - Data
16      Input data set.
17
18Outputs:
19
20   - Reduced Data
21      Data set which selected attributes.
22
23Description
24-----------
25
26Rank widget considers class-labeled data sets (classification or regression)
27and scores the attributes according to their correlation with the
28class.
29
30.. image:: images/Rank-stamped.png
31
321. Attributes (rows) and their scores by different scoring methods
33   (columns).
34#. Scoring techniques and their (optional) parameters.
35#. For scoring techniques that require discrete attributes this is the number
36   of intervals to which continues attributes will be discretized to.
37#. Number of decimals used in reporting the score.
38#. Toggles the bar-based visualisation of the feature scores.
39#. Adds a score table to the current report.
40
41Example: Attribute Ranking and Selection
42----------------------------------------
43
44Below we have used immediately after the :ref:`File`
45widget to reduce the set of data attribute and include only the most
46informative one:
47
48.. image:: images/Rank-Select-Schema.png
49
50Notice how the widget outputs a data set that includes only the best-scored
51attributes:
52
53.. image:: images/Rank-Select-Widgets.png
54
55Example: Feature Subset Selection for Machine Learning
56------------------------------------------------------
57
58Following is a bit more complicated example. In the workflow below we
59first split the data into training and test set. In the upper branch
60the training data passes through the Rank widget to select the most
61informative attributes, while in the lower branch there is no feature
62selection. Both feature selected and original data sets are passed to
63its own :ref:`Test Learners` widget, which develops a
64:ref:`Naive Bayes <Naive Bayes>` classifier and scores it on a test set.
65
66.. image:: images/Rank-and-Test.png
67
68For data sets with many features and naive Bayesian classifier feature
69selection, as shown above, would often yield a better predictive accuracy.
Note: See TracBrowser for help on using the repository browser.