AI-powered employee attrition prediction: Key insights from the Gauteng ICT sector

Résumé

Employee attrition remains a persistent challenge for ICT organisations in Gauteng, South Africa, impacting performance, continuity, and workforce stability. High turnover disrupts project timelines, erodes institutional knowledge, and inflates recruitment and training costs. This study investigates attrition predictors by applying logistic regression to a synthetically generated structured dataset of 1,470 employee records in Gauteng. It examines whether workplace, demographic, and relational factors significantly influence turnover outcomes. Guided by three work-based practice frameworks: Social Exchange Theory, Herzberg’s Two-Factor Theory, and Job Embeddedness Theory, the research aims to identify meaningful retention drivers beyond surface-level attributes. A stratified random sampling method was used to construct a representative subset of 50 records, balancing class proportions within practical constraints. Pre-processing, encoding, and scaling were performed using Python libraries to ensure data integrity, followed by hypothesis testing and regression analysis to evaluate predictor significance. Findings show that relationship satisfaction and organisational commitment negatively correlate with attrition, reinforcing the importance of intrinsic motivators in retention. In contrast, gender, age, tenure, and income-related variables were statistically insignificant, challenging conventional HR assumptions. These results contribute to organisational behaviour literature by highlighting relational and psychological factors over demographic predictors. They suggest that fostering a supportive work environment and cultivating employee engagement may be more effective than relying solely on compensation or tenure-based incentives. The study recommends future research that incorporates larger samples from sectors beyond ICT, employs hybrid methodologies and longitudinal designs, and integrates qualitative insights from interviews or surveys to deepen understanding and inform proactive retention policies.

Citer ce document

Accès au document

Texte intégral en lecture en ligne, réservé aux abonnés SPHAERO et aux membres de l'institution. Se connecter

Voir l'article sur le site de la revue

Auteur(s)

Statistiques

Consultations : 1

Téléchargements : 0