Work

An explainable and fair TabNet framework for remote employee performance prediction using multidimensional workplace data

Year: 2026

Type: article

Abstract: The prediction of performance by remote employees has become more important as organizations try to make their workforce data-driven when distributed. Existing approaches, however, tend to be non-interpretable, non-robust, and non-fair. In this study, an Explainable TabNet Framework for predicting employee performance remotely from a workforce dataset with 100,000 employee records is proposed. This framework incorporates attention-based deep tabular learning, explainability through SHAP values, robustness evaluation via cross-validation, statistical significance testing, and fairness auditing of subgroups in this field. Multiple regression analysis metrics were computed for the predictive performance of the model and compared with four other baselines: Linear Regression, Ridge Regression, Random Forest, Extra Trees, Gradient Boosting, Model XGBoost and LightGBM. The proposed framework outperformed all baseline models with better predictive performance, as evident in the R 2 score, RMSE, MAE, and PCC values, which were 0.9880, 0.1549, 0.1223, and 0.9956, respectively. The most significant predictors of employee performance as determined by ex-ante analysis were employee satisfaction, monthly salary, project involvement, participation in training, and tenure in the organization. Cross-validation and statistical testing ensured model robustness and significance, and fairness testing confirmed that the prediction was similar across groups of demographic and organizational individuals. The Explainable TabNet Framework is an accurate, interpretable, robust, and fair solution for employee performance prediction in remote work, which will enable transparent workforce analytics and human resource decision making based on evidence.

Source: Scientific Reports

DOI: 10.1038/s41598-026-76297-z

Authors: Zihe Zhu, Guangsen Wei, XiaoQian Bian, Sihui Yu

Institutions: New York University, Guangzhou Huashang College, City University of Macau, Shenzhen Polytechnic University

Language: English

Topic: AI and HR Technologies

Subfield: Organizational Behavior and Human Resource Management

Field: Business, Management and Accounting

Domain: Social Sciences

SDGs: Decent work and economic growth, Reduced inequalities

Open Access status: gold

Estimated APC paid: $2,690