A Hybrid Machine Learning Framework for Analyzing Digital Competence Configurations and Task Efficiency

Abstract

The banking industry is undergoing a transformation and digital competence is now a key factor in assessing an employee's performance. In this study, the configurational and predictive relationships between digital competence of bank employees and task efficiency is examined in a dual method analysis, in which fuzzy-set qualitative comparative analysis (fsQCA) and machine learning is applied. A synthetic data was created based on a structure of 500 bank staff members and 55 variables related to the five dimensions of digital competence: information and data competence (IDL), digital communication and collaboration (DCC), digital content creation (DCR), digital safety (DSF), digital problem solving (DPS) and eight task efficiency indicators. The internal consistency of all constructs was found to be good and strongly consistent with reliability analysis (α = 0.798–0.930). Multiple linear regression showed that the total variance in task efficiency was 65.3% (F = 185.6, p < 0.001) accounted for by the digital competence dimensions IDL and DPS with beta of 0.319 and 0.316, respectively. 14 sufficient configurational paths to high task efficiency were found and DPS is a core condition in 10 of these 14 configurations (fsQCA). Logistic regression showed to be the most predictive machine learning classifier (AUC-ROC = 0.851, Accuracy = 0.790) among the 6 classifiers considered. Results show a bi-directionality of digital competence in relation to task efficiency both from the side of the individual and from the side of the task, and provide practical implications for banking HR development and digital transformation strategy.