Abstract
Knowledge is one of the most critical resources that an organisation can possess. However, knowledge only constitutes an asset when it is shared and utilised effectively by firms. This capstone project shows how web applications, such as the R Shiny App, can assist the Human Resource (HR) function apply knowledge and insights from data mining to guide HR initiatives/programmes that can potentially mitigate attrition. Six supervised machine learning (ML) algorithms were applied to the IBM HR dataset to identify the determinants of, and predict, voluntary employee attrition, using R Shiny.
Internal work-related factors such as working overtime, business travel and delayed promotional opportunities were identified as negative determinants of employee attrition. In contrast, behavioural dimensions of Human Capital (HC) were found to be positive determinants of employee attrition. The results showed that the best performing algorithm based on balanced accuracy score was Logistic Regression (AUC = 0.7573). However, Naïve Bayes performed best on the sensitivity metric (Sensitivity = 0.86) while the Decision Tree model achieved the highest specificity score (Specificity = 0.8607). These findings strongly suggested that the choice of ML model for predicting voluntary employee attrition should be guided by a firm’s HR retention strategy (if one such exists) and cost-benefit implications.