Purpose: What do antecedents of turnover tell us when examined using human resources (HR) analytics and machine-learning tools, and what are the respective theoretical and practical implications? Although the turnover literature is expansive, empirical evidence on turnover antecedents studied using data science tools remains limited.Design/methodology/approach: To help reinvigorate research in this field, the authors propose a novel examination of turnover antecedents—competencies, commitment, trust and cultural values—using big data tools to develop a granular, case-dependent measure of turnover.Findings: Using archival data from 700,000 employees of a large organization collected over a period of ten years, the authors find that turnover is generally associated with varying levels of these antecedents. However, in more fine-grained analysis, their relation to turnover is contingent upon role, person and cultural background.Originality/value: The authors discuss the implications on turnover and strategic HR research and the potential of Artificial Intelligence and machine-learning methods in the design and implementation of managerial and HR planning initiatives.

We draw on cognitive and behavioural theories and on the artificial intelligence literature in order to propose a framework of future operator – workstation interaction in the ‘Industry 4.0’ era. We name the proposed framework ‘Operator – Workstation Interaction 4.0’. The latter’s capabilities permit an adaptive, ongoing interaction that aims to improve operator safety, performance, well-being, and satisfaction as well as the factory’s production measures. The framework is composed of three subsystems: (1) the observation subsystem which observes the operator and the processes occurring in the workstation, (2) the analysis subsystem which generates understanding and implications of the observations output, (3) the reaction subsystem which determines if and how to respond. The paper describes these elements and illustrate them using an example of a fatigued worker. The contributions, implications, and limitations of the proposed framework are discussed, and future research directions are presented.

The construction of efficient decision and classification trees is a fundamental task in Big Data analytics which is known to be NP-hard. Accordingly, many greedy heuristics were suggested for the construction of decision-trees, but were found to result in local-optimum solutions. In this work we present the dual information distance (DID) method for efficient construction of decision trees that is computationally attractive, yet relatively robust to noise. The DID heuristic selects features by considering both their immediate contribution to the classification, as well as their future potential effects. It represents the construction of classification trees by finding the shortest paths over a graph of partitions that are defined by the selected features. The DID method takes into account both the orthogonality between the selected partitions, as well as the reduction of uncertainty on the class partition given the selected attributes. We show that the DID method often outperforms popular classifiers, in terms of average depth and classification accuracy.

The negative impact of absenteeism on organizations’ productivity and profitability is well established. To decrease absenteeism, it is imperative to understand its underlying causes and to identify susceptible employee subgroups. Most research studies apply hypotheses testing and regression models to identify features that are correlated with absenteeism—typically, these models are limited to finding simple correlations. We illustrate the use of interpretable classification algorithms for uncovering subgroups of employees with common characteristics and a similar level of absenteeism. This process may assist human resource managers in understanding the underlying reasons for absenteeism, which, in turn, could stimulate measures to decrease it. Our proposed methodology makes use of an objective-based information gain measure in conjunction with an ordinal CART model. Our results indicate that the ordinal CART model outperforms conventional classifiers and, more importantly, identifies patterns in the data that have not been revealed by other models. We demonstrate the importance of interpretability for human resource management through three examples. The main contributions of this research are (1) the development of an information-based ordinal classifier for a published absenteeism dataset and (2) the illustration of an interpretable approach that could be of considerable value in supporting human resource management decision-making.