Data_Preprocessing_Methods_Eng.md

Data Preprocessing Methods

Author: Gaspard-Fauvelle Angel
License: CC BY-NC-SA 2.0

Voici la version française du document Data Preprocessing Methods

Table of Contents

• Feature Engineering
  • Scaling, Normalization, and Standardization
• Encoding Categorical Features
• Data Splitting
• Preprocessing Utilities

---

Preprocessing Utilities

Tool / Concept	Description
fit() / transform() / fit_transform()	Methods used to apply transformations
Pipeline (sklearn)	Chains preprocessing steps and models
ColumnTransformer (sklearn)	Applies transformers to specific columns
FunctionTransformer	Wraps custom functions into a transformer

---

Feature Engineering

Scaling, Normalization and Standardization

Method / Tool (Sklearn only)	Description
MinMaxScaler (sklearn)	Scales features to a 0–1 range
StandardScaler (sklearn)	Standardizes features (mean = 0, std = 1)
RobustScaler (sklearn)	Scales using median and IQR (robust to outliers)
Normalizer (sklearn)	Normalizes samples individually to unit norm

---

Encoding Categorical Features

Method / Tool	Description
OrdinalEncoder (sklearn)	Encodes ordinal features with ordered integers
LabelEncoder (sklearn)	Encode ordinal/ranking categories data
pd.get_dummies() \ OneHotEncoder (sklearn)	Encode nominal categories data

---

Data Splitting

Method / Tool (Sklearn only)	Description
train_test_split()	Splits dataset into train/test subsets (Normal-size dataset splitter)
StratifiedKFold	Maintains class proportions across folds (Bigger-size classification target splitter)
KFold / GroupKFold	Cross-validation methods with or without grouping (Bigger-size regression target splitter)
TimeSeriesSplit	Cross-validation method made for datasets including a date you set as index

---

Utilitaires de Prétraitement

Tools / Concept	Description
fit() / transform() / fit_transform()	Methods for applying transformations
Pipeline (sklearn)	Chains preprocessing and modeling steps
ColumnTransformer (sklearn)	Applies transformers to specific columns
FunctionTransformer	Integrates custom functions into a pipeline

---

You can redirect to the final sheet about End-To-End Data Science Workflow

Edited: 2025/04/22