Employee Attrition Prediction based on Grey Wolf Optimization and Deep Neural Networks

Abstract

Despite the constructive application of promising technologies such as Neural Networks, their potential for predicting human resource management outcomes still needs to be explored. Therefore, the primary aim of this paper is to utilize neural networks and meta-heuristic technologies to predict employee attrition, thereby enhancing prediction model performance. The conventional Grey Wolf optimization optimization (GWO) has gained substantial attention notice because of its attributes of robust convergence, minimal parameters, and simple implementaton. However, it encounter problems with slow convergence rates and susceptibility to local optima in practical optimization scenarios. To address these problems, this paper introduces an enhanced Grey Wolf Optimization algorithm incorporating the utilization of Cauchy-Gaussian mutation, which contributes to enhancing diversity within the leader wolf population and enhances the algorithm's global search capabilities. Additionally, this work preserves exceptional grey wolf individuals through a greedy selection of 2 mechanisms to ensure accelerated convergence. Moreover, an enhanced exploration strategy is suggested to expand the optimization possibilities of the algorithm and improve its convergence speed. The results shows that the proposed model achieved the accuarcy of 97.85%, precision of 98.45%, recall of 98.14%, and f1-score of 97.11%. Nevertheless, this paper extends its scope beyond merely predicting employee attrition probability and activities to enhance the precision of such predictions by constructing an improved model employing a Deep Neural Network (DNN)..