ROLE OF HR ANALYTICS AND PREDICTIVE MODELING IN REDUCING EMPLOYEE ATTRITION

Abstract

Employee attrition poses a significant HRM issue because it costs the organization more to recruit new employees, can lead to a loss of knowledge and skills, can cause a disruption of organizational bench strength and can affect productivity. Retention strategies that are used traditionally are rooted in historic information on turnover, exit interviews, and the intuition of managers while on the job. All of these have their value but are less effective at predicting those who might leave before they do. HR analytics and predictive modelling is a data driven solution that can take employee demographic, payroll, performance, engagement, career growth, workload and organisational data into account to predict the likelihood of employee attrition and why, greatly. Over the past few years, the research done in this field demonstrated that some of the machine learning techniques employed for employee turnover prediction are logistic regression, decision trees, random forest, support vector machine, gradient boosting and ensemble learning. A systematic review of 52 studies revealed that supervised learning is the dominating method in the literature and that the majority of predictors of turnover mentioned in the literature were salary and overtime. Subsequent studies have incorporated explainable AI (XAI) and the use of AI interpretation in particular with SHAP into predictive accuracy and action taken by HR.

This research investigates the impact of HR analytics and predictive modelling on employee retention by a conceptual and empirical framework that integrates these two concepts. The research design is quantitative and analytical with an illustrative data of employees that have 1500 observations and 10 variables relating to compensation, tenure, job satisfaction, overtime, performance, promotion, workload, engagement, and career development. Using accuracy, precision, recall, F1-score and ROC-AUC all of the aforementioned models are compared with logistic regression model, decision tree model, random forest model, support vector machine and gradient boosting model. The illustrative results show that both gradient boosting and random forest give superior predictive performance compared to classical logistic regression and that job satisfaction, overtime, career progression, income, workload and organizational engagement are significant predictors. The analysis also shows how predictive analytics can help differentiate individuals to enable targeted retention interventions for those deemed to be at risk, from those who are not. When considered as predictive technologies, HR analytics is more valuable when combined with the explanatory power, the governance of the data, managerial input, and continuous review.

The major problem for human resource management is employee attrition, since it brings high costs related to the loss of productivity.Employee attrition is of great concern as it leads to significant losses due to diminished productivity.