Orange Forum • View topic - Building Histograms for attribute lists

Building Histograms for attribute lists

A place to ask questions about methods in Orange and how they are used and other general support.

Building Histograms for attribute lists

Postby veeraganeshreddy » Mon Aug 13, 2007 15:03

Hi i got 2 questions

how to build histograms for attribute lists using orange and how to construct attribute lists for the attributes in a data set, where attribute list contains attribute values, corresponding class and a index for each value so that we can sort the attribute list in ascending order.

Is it possible to find gini index for groups of attribute lists, Say i got 9 values in the 1 attribute list and i group them to 3 where each group contains 3 values.

Also is there any IRC channel for Orange?

Postby Blaz » Wed Aug 15, 2007 15:06

Answering this may be difficult, as I have to admit I do not understand your questions well. But let's try:

how to build histograms for attribute lists using orange and how to construct attribute lists for the attributes in a data set, where attribute list contains attribute values, corresponding class ...


Would attribute lists be something that is standardly refered to as a "data instance" or "an example" in machine learning. If so, then is so what Orange does. A collection of these is called a data set.

For displaying a histogram of a continuous variable, you may use something like:

Code: Select all
import orange
import pylab

data = orange.ExampleTable("iris")
pylab.hist([float(d["sepal length"]) for d in data])
pylab.show()


or use Distributions from Visualization pane in Orange Canvas.

and a index for each value so that we can sort the attribute list in ascending order.


Instead of an index, name your attribute values in the second line of your tab-delimited data file (write an attribute value list instead of "discrete").

Is it possible to find gini index for groups of attribute lists, Say i got 9 values in the 1 attribute list and i group them to 3 where each group contains 3 values.


Gini index was designed to score one attribute, if you would like to score a group, you would need to construct a new attribute (e.g., constructive induction) from the group and score it.

Also is there any IRC channel for Orange?


No.


Return to Questions & Support



cron