Awesome Feature Engineering Project Introduction

Project Overview

Awesome Feature Engineering is a curated list specifically collecting machine learning feature engineering technical resources. The project is maintained by Andrei Khobnia and follows the Creative Commons Attribution-Noncommercial-ShareAlike 3.0 Unported License.

This project provides a comprehensive repository of feature engineering techniques for machine learning practitioners, covering methods and tools for various data types.

Main Content Categories

1. Numeric Data

• Data Transformation:

  • Box-Cox Transformation: scipy.stats.boxcox
  • Log Transformation: np.log (x + const)

• Automated Feature Engineering:

  • Featuretools: For automated feature engineering

• Feature Interaction:

  • sklearn.preprocessing.PolynomialFeatures: Polynomial feature generation
  • Division operations
  • Other interactive features

2. Textual Data

• Bag-of-words Model:

  • Bag-of-words model
  • A Gentle Introduction to the Bag-of-Words Model
  • sklearn.feature_extraction.text.CountVectorizer
  • sklearn.feature_extraction.DictVectorizer
  • sklearn.feature_extraction.FeatureHasher

• Word Embedding Techniques:

  • Word embedding
  • GloVe: Global Vectors for Word Representation
  • Gensim: models.word2vec – Word2vec embeddings
  • fastText
  • Word2Vec and FastText Word Embedding with Gensim

• Feature Extraction Techniques:

  • ClearTK - Feature Extraction Tutorial
  • Regular Expressions
  • Part-of-Speech_Tagging
  • NLTK Categorizing and Tagging Words

3. Image Data

• Traditional Feature Extraction:

  • Feature extraction and similar image search with OpenCV for newbies
  • OpenCV -- Feature Detection and Description
  • SimpleCV.Features package
  • Scikit-image feature module

• Deep Learning Feature Extraction:

  • Keras pre-trained models feature extraction
  • Using Keras' Pre-trained Models for Feature Extraction in Image Clustering

4. Categorical Data

• One-Hot Encoding:

  • Why One-Hot Encode Data in Machine Learning?
  • How to One Hot Encode Sequence Data in Python
  • sklearn.preprocessing.OneHotEncoder
  • Keras - to_categorical

• Target Encoding:

  • Likelihood encoding of categorical features
  • Python target encoding for categorical features
  • Adding variance column when mean encoding

• Feature Hashing:

  • Feature Hashing on Wikipedia
  • Feature hashing and Extraction in VowpalWabbit
  • Feature hashing in scikit-learn

5. Time Series Data

• Automatic Feature Extraction:
  • Automatic extraction of relevant features from time series
  • Basic Feature Engineering With Time Series Data in Python

6. Geospatial Data

• Includes feature engineering techniques related to geographical location.

Project Features

1. Comprehensiveness: Covers major data types in machine learning and their corresponding feature engineering techniques.
2. Practicality: Provides specific tool libraries and code implementations.
3. Open Source: Adopts an open-source license, welcoming community contributions.
4. Authoritativeness: Links to authoritative documentation, tutorials, and academic resources.
5. Actionability: Offers specific Python libraries and function call methods.

Value Proposition

This project is particularly valuable for the following groups:

• Machine Learning Engineers
• Data Scientists
• Feature Engineering Researchers
• Machine Learning Beginners
• Practitioners looking to improve model performance

Contribution Methods

The project encourages community contributions; new resources can be added or existing content improved by creating pull requests.

Summary

The Awesome Feature Engineering project provides a comprehensive and practical resource library for machine learning feature engineering, serving as an important reference for learning and applying feature engineering techniques. Through systematic categorization and rich resource links, it helps practitioners quickly find suitable feature engineering methods for specific data types.