Show HN: Automate Your Machine Learning Experiments: try mljar-supervised

Show HN: Automate Your Machine Learning Experiments: try mljar-supervised

Build Status
PyPI version
PyPI pyversions

Computerized Machine Learning

mljar-supervised is an Computerized Machine Learning python package. It goes to organize ML fashions for:

  • binary classification,
  • multi-class classification,
  • regression.

What’s participating in it?

  • mljar-supervised creates markdown experiences from AutoML working in opposition to. The instance of AutoML leaderboard abstract:

AutoML leaderboard

The instance for Decision Tree abstract:
Decision Tree summary

The instance for LightGBM abstract:
Decision Tree summary

  • This package is computing Baseline in your information. So you’ll know in disclose for you Machine Learning or not! It’s possible you’ll maybe effectively perchance understand how participating are your ML fashions evaluating to the Baseline. The Baseline is computed in keeping with prior class distribution for classification, and simple imply for regression.
  • This package is working in opposition to simple Decision Bushes with max_depth <= 5, as a plan to additionally with out hazard visualize them with unbelievable dtreeviz to higher hint your information.
  • The mljar-supervised is using simple linear regression and embrace its coefficients within the abstract file, as a plan to additionally confirm which points are extinct essentially the most within the linear mannequin.
  • It's a great distance using an limitless residence of algorithms: Random Forest, Additional Bushes, LightGBM, Xgboost, CatBoost (Neural Networks will probably be added quickly).
  • It goes to pause points preprocessing, take pleasure in: lacking values imputation and changing categoricals. What's extra, it's going to additionally deal with goal values preprocessing (It's possible you'll maybe effectively perchance not trust how usually it is miles wished!). As an illustration, changing particular goal into numeric.
  • It goes to tune hyper-parameters with no longer-so-random-search algorithm (random-search over outlined residence of values) and hill climbing to beautiful-tune last fashions.
  • It goes to compute Ensemble in keeping with greedy algorithm from Caruana paper.
  • It cares about explainability of fashions: for each algorithm, the attribute significance is computed in keeping with permutation. Furthermore, for each algorithm the SHAP explanations are computed: attribute significance, dependence plots, and determination plots (explanations might maybe effectively perchance be switched off with explain_level parameter).

Fast instance

There might maybe be a simple interface available with match and predict programs.

import pandas as pd
from supervised.automl import AutoML

df = pd.read_csv("", skipinitialspace=Truthful)

X = df[df.columns[:-1]]
y = df["income"]

automl = AutoML(results_path="directory_with_reports")
automl.match(X, y)

predictions = automl.predict(X)

For essential components please confirm AutoML API Scientific doctors.



From PyPi repository:

pip set up mljar-supervised

From provide code:

git clone
cd mljar-supervised
python set up

Arrange for developing

git clone
virtualenv venv --python=python3.6
provide venv/bin/urged
pip set up -r necessities.txt
pip set up -r requirements_dev.txt


The mljar-supervised is an delivery-supply undertaking created by MLJAR. We care about ease of use within the Machine Learning.
The presents a good and simple consumer interface for developing machine learning fashions.


Please enter your comment!
Please enter your name here