TailBooster framework enhances AI prediction of extreme events

Researchers have developed TailBooster, a novel dual-layer generative framework designed to improve machine learning models by augmenting rare extreme events in tabular data. The framework addresses limitations in conventional generative models by focusing on distributional tails and ensuring operational validity of synthetic data. TailBooster combines a statistical layer for extreme value extraction with a deep learning layer for anomaly detection, discarding infeasible instances. Evaluations on US flight records demonstrated significant improvements in predicting extreme air times and arrival delays, reducing Mean Absolute Error by up to 57% compared to standard synthetic data augmentation. AI

IMPACT Enhances AI model performance in predicting rare, critical events by improving synthetic data generation.

RANK_REASON The cluster contains a research paper detailing a new framework for data augmentation. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• mean absolute error
• Tabular Variational Autoencoder
• TailBooster
• US
• variational auto-encoder

• paper
• model release

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