pycaret beta 0.0.6 has been released. We have made fundamental progress in last few days. Some of the key achievements and updates are as follows:
1. NEW MODULES
==============
- pycaret.nlp is now publicly available. pycaret's NLP module is one of it's kind for combining supervised and unsupervised machine learning. It supports more than 6 different kind of topic models along with library of 40+ estimators to choose from (when learning is set to semi-supervised). plot_model function in NLP is completely interactive and makes the best use of plotly interface in python.
-pycaret.datasets is now created. It's a big data respository hosted as github.com/pycaret/pycaret. This will allows to directly get data from our git without ever downloading and storing csv files. Example:
from pycaret.datasets import get_data
data = get_data('credit')
to view index, please see get_data('index')
2. XGBOOST AND LIGHTGBM NOW AVAILABLE
=====================================
- xgboost and Microsoft's lightgbm is now generally available in classification and regression modules. For beta release 0.0.7 and 0.0.8 we are hoping catboost will be included.
3. PRE-PROCESS MODULE
=====================
- Great progress has been made for pre-processing module. Will be release in beta 0.0.9
4. BUG FIX
==========
With our first major release of 0.0.4, couple of bugs have been identified which are fixed now.
- AutoML regression was ignoring boosting models - FIXED
- AutoML classification boosting conflict - FIXED
- pycaret.classification setup sampling value fixed to 25000 from 20000 in previous version.
- documentation added for evaluate_model(), save_model(), load_model(), save_experiment(), load_experiment()
5. DOCUMENTATION
================
- README documentation is now updated at https://github.com/pycaret/pycaret
- Web development started. Targetted for 2nd week of January 2020.
6. DEFAULT CHANGES
=================
- Default metric in Regression is now set to R2 from 'MAE' in previous versions.
