LLMs drive automated feature engineering for structured data

Researchers have developed Evolutionary Feature Engineering (EFE), a novel framework that leverages large language models (LLMs) to automatically discover preprocessing transformations for structured data. EFE represents these transformations as Python programs, enabling seamless integration into existing machine learning pipelines. The framework refines candidate programs using dataset context, summary statistics, and downstream performance feedback. EFE has demonstrated success in time-series forecasting, reducing errors by 3% or more with models like Chronos-2, and in tabular prediction, where it evolves compact, interpretable feature programs that match or exceed existing LLM-based methods. AI

IMPACT Automates complex data preprocessing, potentially improving accuracy and interpretability of ML models across various domains.

RANK_REASON The cluster describes a new research paper detailing a novel framework for feature engineering using LLMs. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• Chronos 2 Forecasting Model
• EFE-Tab
• EFE-Time
• Evolutionary Feature Engineering
• Hugging Face
• large-language models
• Mae
• Mase
• Python
• WQL

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →