Rankeval

Starting with this release, RankEval is available from the Python Package Index (PyPI). You may easily install the tool by running:

pip install rankeval

Python 3 support is in progress, still not stable. Improved documention and the notebooks showing tool functionalities.

Starting with this release, RankEval is fully available from the Python Package Index (PyPI) and the code is packaged into several binary wheels according to the most adopted platforms (linux and osx, python versions 2.7, 3.4, 3.5, 3.6 and 3.7). You may easily install the tool by running:

pip install rankeval

and most probably you don't even need cython locally to compile low level code since the binaries should already be available for your platform.

Another big improvement is the introduction of Continuous Integration using Travis. Each commit/release/PR is now tested before having the possibility to be merged into the repo. Even releasing the library on pypi has been automatized: each tagged commit on the master will now activate the deployment actions that will test again the code, package and release it for the several platforms it will support.

Starting from this release, RankEval now officially supports python3.

We added also two additional model proxies, namely CatBoost and Jforests.

Finally, we added a feature for plotting the shape of a single tree in the ensemble, with the possibility to personalize it in several ways. We modified the topological notebook accordingly, reporting how to use this new feature.