Employee performance evaluation has always depended on subjective evaluations, which are subject to inconsistency and presumption. This study minimises the problem by introducing an AI-driven behavioural analytics model that utilises both structured data and unstructured feedback to deliver a more impartial and insightful evaluation system. The model integrates advanced feature engineering with a hybrid architecture of XGBoost and BERT, allowing it to understand numerical patterns and derive significant insights from natural language. Feature importance ratings and SHAP interpretations enhance the transparency of prediction outcomes, providing the system appropriate for practical HR applications. Empirical evaluations utilising a synthesised workplace dataset demonstrated enhanced performance, achieving a R2 score of 0.942, an F1 score of 0.91, and an overall prediction accuracy of 94.8%, significantly outperforming baseline models such as Linear Regression, Decision Trees, and SVR. The model predicts performance with high accuracy and provides behavioural insights that assist HR departments in identifying developmental needs and objectively recognising outstanding performers. This approach combines explainable AI with principles of behavioural psychology, resulting in a more ethical and intelligent evaluation framework compared to previous studies. This study connects conventional evaluations with data-informed decision-making in human resource management.
Paper
Full text
AI Based Behavioural Analytics Model for Workplace Performance Evaluations
Semantic Scholar · 2025
Abstract
Employee performance evaluation has always depended on subjective evaluations, which are subject to inconsistency and presumption. This study minimises the problem by introducing an AI-driven behavioural analytics model that utilises both structured data and unstructured feedback to deliver a more impartial and insightful evaluation system. The model integrates advanced feature engineering with a hybrid architecture of XGBoost and BERT, allowing it to understand numerical patterns and derive significant insights from natural language. Feature importance ratings and SHAP interpretations enhance the transparency of prediction outcomes, providing the system appropriate for practical HR applications. Empirical evaluations utilising a synthesised workplace dataset demonstrated enhanced performance, achieving a R2 score of 0.942, an F1 score of 0.91, and an overall prediction accuracy of 94.8%, significantly outperforming baseline models such as Linear Regression, Decision Trees, and SVR. The model predicts performance with high accuracy and provides behavioural insights that assist HR departments in identifying developmental needs and objectively recognising outstanding performers. This approach combines explainable AI with principles of behavioural psychology, resulting in a more ethical and intelligent evaluation framework compared to previous studies. This study connects conventional evaluations with data-informed decision-making in human resource management.