Sakhawat Hussain, Tanim and Rahanuma, Tarannum and Md Manarat Uddin, Mithun (2023) Privacy-Preserving Behavior Analytics for Workforce Retention Approach. American Journal of Engineering , Mechanics and Architecture, 1 (9). pp. 188-215. ISSN 2993-2637

Text 188-215+Privacy-Preserving+Behavior+Analytics+for+Workforce.pdf Download (994kB)

Abstract

Attrition among employees is a paramount issue to contemporary organizations which directly affects the continuous running of activities, employee morale and recruitment or training expenses of new members of staff. Behavioral analytics is also becoming relevant in the Human Resource (HR) departments, where it is used to detect early signs of dissatisfaction in employees and foretell attrition. The classic implementations of workforce analytics tend to invade the privacy of employees by revealing their personal and behavioral sensitive information. This study focuses on the twin challenge of predictive performance and data privacy and proposes a privacy-guaranteed architecture of behavioral analytics in support of workforce retention initiatives. This paper accesses and helps discern two major factors that impact employee turnover using the publicly available HR Employee Attrition dataset, analyzing them accordingly. Such factors as job role, business travel frequency, overtime hours, distance from home, and years at company are investigated to reveal their correlation with attrition. The aim is to develop precise predictive models of employee retention without ever touching or revealing raw individual data. In order to maintain privacy, the study adds differential privacy and federated learning to the machine learning pipeline. Differential privacy adds statistical noise to sensitive variables, allowing them not to be re-identified about individuals, but preserving the overall utility of data. Federated learning emulates decentralized training of a model on different departments or job titles, such that collaborative analytics can be achieved without centralized data exchange. Using these methods, the research illustrates how organizations may gain insightful knowledge, at the same time keeping the data of individuals anonymous and secure. The findings indicate that, although privacy-preserving methods imply certain trade-offs in model performance, they substantially improve the ethical considerations and employee confidence in data governance procedures. Feature importance analysis shows the most significant behavioral attributes of attrition, which provides HR managers with intelligent action points to apply specific retention efforts.This study is relevant to the emerging ethical AI in workforce management since it shows it is possible to have effective retention analytics without invading the privacy of employees. This study explores the implementation of privacy-enhancing technologies on HR systems and praises a future wherein predictive workforce planning and data protection principles can co-exist. Attrition can be managed proactively by the organizations that will adopt such a framework in a way that is transparent and respectful of the individual privacy rights.

Item Type:	Article
Subjects:	T Technology > TJ Mechanical engineering and machinery
Divisions:	Postgraduate > Master's of Islamic Education
Depositing User:	Journal Editor
Date Deposited:	22 Sep 2025 05:16
Last Modified:	22 Sep 2025 05:16
URI:	http://eprints.umsida.ac.id/id/eprint/16368

Actions (login required)

View Item