Data and feature engineering · MLD-0263
Data leakage
Use of information during training or selection that would not be available for the intended prediction.
Original reference page · detailed expansion pending. Original definitions, examples, notes and citations are preserved below. New examples are illustrative, not benchmark results.
Original example
Scale features using the entire dataset before splitting.
Original caveat
Leakage can arise from preprocessing, duplicates, or labels.
Context and aliases
Level: Intermediate. Stages: Data · Training · Evaluation.
Tags: ML · DL · SLM · LLM.
Aliases: Information leakage.
Related concepts
Concepts linking here
Technology connections
- Argilla — Human feedback and dataset-curation platform.
- Cleanlab — Data-centric tools for finding label and other dataset issues.
- CVAT — Computer-vision annotation tool.
- Docling — Document parsing and conversion toolkit.
- FiftyOne — Visual dataset exploration and evaluation application.
- Great Expectations — Data validation framework using declarative expectations.
- Label Studio — Configurable data-labeling application.
- Pandera — DataFrame schema and statistical validation library.
- TensorFlow Data Validation — Dataset statistics and anomaly-detection tooling.
- Unstructured — Document ingestion and partitioning toolkit.
Original references
- scikit-learn: cross-validation ↗ · Reference link
- scikit-learn: missing-value imputation ↗ · Reference link
- scikit-learn: preprocessing ↗ · Reference link
- scikit-learn: feature selection ↗ · Reference link
- Original reference URL ↗