python-weka-wrapper3 on mloss.org
I finally managed to create a project page for my python-weka-wrapper3 project on mloss.org: http://mloss.org/software/view/660/
Open-Source and related Stuff
I finally managed to create a project page for my python-weka-wrapper3 project on mloss.org: http://mloss.org/software/view/660/
Since there have been a number of little improvements, I decided to make a new release: python-weka-wrapper: 0.3.9 python-weka-wrapper3: 0.1.1 Here are the changes: plot_learning_curve method of module weka.plot.classifiers now accepts a list of test sets; * is index of…
Read more
Just hacked together a simple library for reading Quicken Interchange Format (QIF) files. Called it quicken4j and made it available on Github and Maven Central: https://github.com/fracpete/quicken4j
Had some bug-reports in regards to how the jcamp-dx library was (mis)handling NMR data. Since these have been all addressed, I made a new release today: 0.9.4
Well, despite Azure offering Linux in their cloud, the Linux support for signing up is lacking. If you’re on Linux (Firefox or Chromium) you get the following screen which is stuck in an endless-loop: Turns out, you have to be…
Read more
Just created a new github repository that lists all my repositories on my github account (meta-repository??): https://github.com/fracpete/projects Reason for this was to have a sorted list of production-ready repositories rather than the lists offered by github: either all repositories sorted…
Read more
Spent today extending the missing-values-imputation Weka package and added the following algorithms: MultiImputation SimpleNearestNeighbor UserSuppliedValues IRMI based on Chris Beckham’s code Still a bit alpha, so bear that in mind when running experiments. 😉
Weka only has its ReplaceMissingValues filter for dealing with missing values, which just uses means and modes as replacement. However, this approach is not always the best. Hence I started on a Weka package project some time ago to offer…
Read more
Anyone, who ever had to deal with Weka‘s option handling, knows that it is extremely error prone and cumbersome. The main reason is the separate handling of options through command-line and GUI. This results in duplicate text for options, different…
Read more
So, far the multisearch package didn’t allow you to specify what class label the metric (like F1, recall or precision) was optimizing, it was hardcoded to be the first one. With the 2016.6.6 release of the package, the MultiSearch classifier…
Read more