Bugfix release for python-weka-wrapper: 0.1.17

Of course, as soon as you make a release, someone will spot a bug the day after… :-)

Posted on December 18, 2014 at 9:50 am by fracpete · Permalink · Leave a comment
Tagged with: , , ,

Bugfix release for python-weka-wrapper: 0.1.14

Released version 0.1.14 of the python-weka-wrapper library, which consists mainly of a fixed installer and detailed install instructions.
Just in time for Xmas. ;-)

Posted on December 16, 2014 at 2:06 pm by fracpete · Permalink · Leave a comment
Tagged with: , , ,

multisearch weka package released

A long time ago, I added a Weka meta-classifier for parameter optimization called MultiSearch. In contrast to GridSearch, which forces you to optimize two parameters (hence grid), this scheme allows you to optimize an arbitrary number of parameters. However, it does not allow you to extend the search space like GridSearch.

Today, I moved this code into a separate project on github and turned it into a Weka package, making it easier for other people to use it as well.

You can find the code and package archives here:
https://github.com/fracpete/multisearch-weka-package

Posted on December 10, 2014 at 4:38 pm by fracpete · Permalink · Leave a comment
Tagged with: , , ,

python-weka-wrapper 0.1.13 released

Release with minor improvements:

Posted on November 1, 2014 at 8:02 pm by fracpete · Permalink · Leave a comment
Tagged with: , , ,

python-weka-wrapper 0.1.12 released

Today, I made a new release of the Python wrapper for Weka available.
Here is an overview of the changes:

Posted on October 17, 2014 at 11:56 am by fracpete · Permalink · Leave a comment
Tagged with: , ,

confusionmatrix package release

Just released a new version of the confusionmatrix Weka package that I started a while ago. I added a new heatmap visualization, which scale the rows by the sum of counts in that row. Essentially using percentages in a row, which is useful when you have skewed class distributions. Then the big class(es) don’t drown out the smaller (and potentially very important) class(es).