Dynamic and Adaptive Feature Generation with LLM

• DOI: 10.48550/arxiv.2406.03505
• Publication Date: 2024-06-04
• Authors: Xinhao Zhang; Jinghan Zhang; Banafsheh Rekabdar; Yuanchun Zhou; Pengfei Wang; Kunpeng Liu
• Keywords: Feature (linguistics)

The representation of feature space is a crucial environment where data points get vectorized and embedded for upcoming modeling. Thus the efficacy machine learning (ML) algorithms closely related to quality engineering. As one most important techniques, generation transforms raw into an optimized conducive model training further refines space. Despite advancements in automated engineering generation, current methodologies often suffer from three fundamental issues: lack explainability, limited applicability, inflexible strategy. These shortcomings frequently hinder limit deployment ML models across varied scenarios. Our research introduces novel approach adopting large language (LLMs) feature-generating prompts address these challenges. We propose dynamic adaptive method that enhances interpretability process. broadens applicability various types tasks draws advantages over strategic flexibility. A broad range experiments showcases our significantly superior existing methods.