TabFSBench: Tabular Benchmark for Feature Shifts in Open Environments

TL;DR

This paper introduces TabFSBench, a comprehensive benchmark for evaluating the impact of feature shifts in open environments on tabular data models, highlighting their limited robustness and the linear relationship between feature importance and performance degradation.

Contribution

It presents the first systematic study and benchmark for feature shifts in tabular data, including evaluation of large language models and insights into model robustness.

Findings

Most tabular models have limited applicability under feature shifts.

Feature importance shifts linearly relate to performance degradation.

Model performance in closed environments correlates with performance under feature shifts.

Abstract

Tabular data is widely utilized in various machine learning tasks. Current tabular learning research predominantly focuses on closed environments, while in real-world applications, open environments are often encountered, where distribution and feature shifts occur, leading to significant degradation in model performance. Previous research has primarily concentrated on mitigating distribution shifts, whereas feature shifts, a distinctive and unexplored challenge of tabular data, have garnered limited attention. To this end, this paper conducts the first comprehensive study on feature shifts in tabular data and introduces the first tabular feature-shift benchmark (TabFSBench). TabFSBench evaluates impacts of four distinct feature-shift scenarios on four tabular model categories across various datasets and assesses the performance of large language models (LLMs) and tabular LLMs in the tabular benchmark for the first time. Our study demonstrates three main observations: (1) most tabular models have the limited applicability in feature-shift scenarios; (2) the shifted feature set importance has a linear relationship with model performance degradation; (3) model performance in closed environments correlates with feature-shift performance. Future research direction is also explored for each observation. Benchmark: https://github.com/LAMDASZ-ML/TabFSBench.