Add a modular preprocessing module for data preparation

[iamprasadraju]

Description

Introduce a trueml.preprocessing module that provides essential data preprocessing utilities for machine learning workflows. The module should be internally modular while exposing a simple, unified public API through trueml.preprocessing.

The goal is to provide lightweight, educational implementations that integrate naturally with the rest of TrueML.

Goals

• Create the trueml.preprocessing package.
• Design a modular internal package structure.
• Support both NumPy arrays and Pandas DataFrames where appropriate.
• Keep implementations simple, readable, and educational.
• Expose a flat public API through trueml.preprocessing.
• Maintain a consistent API across all preprocessing utilities.

Proposed Package Structure

trueml/
└── preprocessing/
    ├── __init__.py
    ├── scaling/
    │   ├── __init__.py
    │   ├── standard.py
    │   ├── minmax.py
    │   ├── robust.py
    │   ├── maxabs.py
    │   └── normalizer.py
    ├── encoding/
    │   ├── __init__.py
    │   ├── label.py
    │   ├── onehot.py
    │   └── ordinal.py
    ├── imputation/
    │   ├── __init__.py
    │   ├── mean.py
    │   ├── median.py
    │   ├── most_frequent.py
    │   └── constant.py
    ├── transformation/
    │   ├── __init__.py
    │   ├── log.py
    │   ├── polynomial.py
    │   └── power.py
    ├── feature_selection/
    │   ├── __init__.py
    │   ├── variance_threshold.py
    │   ├── correlation.py
    │   └── select_k_best.py
    ├── decomposition/
    │   ├── __init__.py
    │   └── pca.py
    ├── sampling/
    │   ├── __init__.py
    │   ├── oversampling.py
    │   └── undersampling.py
    └── splitting.py

Public API

from trueml.preprocessing import (
    StandardScaler,
    MinMaxScaler,
    RobustScaler,
    LabelEncoder,
    OneHotEncoder,
    MeanImputer,
    train_test_split,
)

Tasks

Scaling

• StandardScaler
• MinMaxScaler
• RobustScaler
• MaxAbsScaler
• Normalizer

Encoding

• LabelEncoder
• OneHotEncoder
• OrdinalEncoder

Missing Value Imputation

• MeanImputer
• MedianImputer
• MostFrequentImputer
• ConstantImputer

Data Splitting

• train_test_split

Feature Transformation

• LogTransformer
• PolynomialFeatures
• PowerTransformer

Feature Selection

• VarianceThreshold
• CorrelationFilter
• SelectKBest

Dimensionality Reduction

• PCA

Sampling

• RandomOverSampler
• RandomUnderSampler

Future Enhancements

• KNNImputer
• TargetEncoder
• FeatureHasher
• SMOTE
• ADASYN
• Recursive Feature Elimination (RFE)
• Mutual Information feature selection
• Column-wise transformations
• Preprocessing pipelines

Acceptance Criteria

• Modular internal architecture.
• Flat public API via trueml.preprocessing.
• Consistent API across all preprocessing components.
• NumPy-first implementation with optional Pandas support.
• Comprehensive documentation with examples.
• Unit tests for every preprocessing component.