Exploring Digital Competence Configurations and Attention Performance Through fsQCA and Machine Learning

Abstract

Digital competence among bank employees is an increasingly critical determinant of task efficiency in technology-driven financial service environments. In this study, a hybrid analytical approach is used that combines Fuzzy-Set Qualitative Comparative Analysis (fsQCA) with ten machine learning (ML) classifiers. The combination of these two approaches is designed to predict the Focused vs. Distracted attention status of bank employees based on multimodal behavioural and digital interaction features during task execution. The dataset included 1,000 observations of bank employees, capturing head orientation, facial expression, gaze direction, micro-movements, application usage, notification interruptions, and task completion time during digital work sessions. There was a marked imbalance in the class distribution with 91.2% of bank employees categorized as Distracted and 8.8% of bank employees categorized as Focused, which was consistent with the actual prevalence of digital distraction in professional banking environments. The LightGBM model performed best with all the classifiers, with an accuracy score of 98.9%, an AUC of 0.9998, and an F1 score of 0.989. Random Forest and XGBoost followed closely with an AUC of 0.9996 and 0.9995, respectively. The top 3 explainable features identified by SHAP analysis were Head Orientation, Facial Expression and Micro-Movements. The necessity analysis using the fsQCA method confirmed that forward gaze (consistency = 0.97), straight head posture (0.94) and neutral facial expression (0.99) are almost necessary conditions for focused attention. The results provide actionable insights for intelligent workforce management systems, adaptive digital workplace platforms, and real-time employee performance monitoring systems in banking and financial service organizations.