Talent management by predicting employee attrition using enhanced weighted forest optimization algorithm with improved random forest classifier

Abstract

Predictive analysis has been an important field of research suitable for a wide range of applications covering a huge volume of domains in predicting the future with the current and past data. In an organisation, the predicted insights are highly helpful in analysing all the aspects of an issue and making decisions suitably. More specifically, talent management requires making an appropriate decision in employing and maintaining suitable skills in the appropriate place. Machine learning algorithms are most commonly used in analysing the attributes that affect employee attrition and predicting employee turnover. This paper presents the prediction model that makes use of an enhanced weight-based forest optimization algorithm. It employs mutual information for selecting the significant features and a modified random forest for classifying the attrition results. The experimental analysis has been performed with the International Business Machines (IBM) human resource employee attrition dataset and the results are compared with the other existing models. The analysis shows that the proposed model offers better results with an accuracy of 91.23% and a minimum error rate of 8.77% than several other models. The feature significance helps in making effective steps in retaining the talents for the benefit of the organization.