ML Helper

Helpers to speed up and structure machine learning projects.

The library is available in Pypi

Installing

The easiest way to install ml-helper is through pip

pip install ml-helper

To use it in your project, you must first import the library

from ml_helper.helper import Helper

And then create a Helper object with a dictionary of keys related to your project

KEYS = {
    'SEED': 1,
    'TARGET': 'y',
    'METRIC': 'r2',
    'TIMESERIES': True,
    'SPLITS': 5
}

hp = Helper(KEYS)

After this, you may use the helper object's many functions

Dependencies

ML-Helper requires:

• Python (>3.5)
• Numpy (>=1.16)
• Pandas (>=0.23.4)
• Seaborn (>=0.9)
• Scikit-learn (>=0.20)
• Natplotlib (>=3)
• Scipy (>=1)
• Imblearn
• Vecstack

Functionality

The functionality is separated into 4 groups:

• Data Exploration
  • Missing Data
  • Boxplot of numerical variables
  • Coefficient of variation
  • Correlation (numerical and categorical)
  • Under Represented Features
  • Target Variable Distribution
  • Feature Importance
  • PCA Component Variance
• Data Preparation
  • Convert features to categories
  • Drop multiple columns
• Modeling
  • Cross Validation (with stratified kfolds, or time series split depending on use case)
    • Randomized Grid Search
  • Pipeline: Collection of models and pipeline steps that get performed and scored
  • Predict: Predict on unseen data
  • Stack Predict: Build a stacked model and perform a prediction
• Regression
  • Plots for predictions
• Classification
  • ROC Curve
  • Classification Report
• Others
  • Select features based on types
  • Split X and y
  • Plot models/pipelines

Working Examples

If you wish to see the library in use, you may view the notebooks in the examples section.

Also, you can see the implementation in their corresponding Kaggle Kernels:

• Bike Sharing in Washington D.C.: Time Series Regression

• Employee Attrition: Classification

ML-Helper Coding Style

Ml-Helper complies to PEP8 and uses black for coding standards

Versioning

SemVer is used for versioning.

License

This project is licensed under the MIT License - see the License file for details