Article Preview
TopIntroduction
The world is a land of opportunities, and talent and skills are abundant among people from different walks of life. This talent is found in all sections and strata of the society irrespective of age and demographical variations. Recruiting such talent in organizations is important, but it is equally essential to satisfy and retain them since employees have their considerations for choosing and working in an organization, and if they are not satisfied, they can resign. This will lead to employee attrition and, thus, a phenomenon is known as employee churn. The reasons for this churn (Cheng et al.,2019) are varied, and there is a need to identify and target prospective employees who are more likely to resign. Therefore, there is a need to forecast churn (also called ‘Turnover_status’) to retain organizations' talent. All people today have their own needs and wants according to the Maslow Hierarchy of Needs (Maslow,1943), and they make real-time decisions to stay in an organization or not based on many factors. According to Hertzberg Two Factor (Motivation and Hygiene) Theory (Herzberg,2005), they need incentives and motivation to work in the organization since they drive the employee's passion towards his work. Additionally, every employee feels a sense of ownership about his work, and only if his contribution to the organization is valued and rewarded will he be satisfied according to the Theory of Organizational Equilibrium (Simon and March,1976).
Thus, every employee needs to be motivated and perceive his job as meaningful according to the Job Embeddedness Theory (Mitchell et al.,2001); otherwise, he may resign from the organization, causing job churn.
To alleviate this problem, there is a need to predict accurately churn with the key factors that influence this decision using an employee’s customized data available online. This is accomplished using state-of-the-art machine learning techniques that consider churn reasons as input in their model and are trained based on past employee data to recognize patterns and identify prospective employees who are likely to quit the organization.
There are existing studies that attempt to predict employee churn (Sisodia et al., 2017; Ogbonnaya et al., 2017; Ma et al., 2019; Keegan and Hartog, 2019) using the above machine learning techniques but a multi-factor hybrid approach for validating the reasons for employee churn was not adopted. Further, the models are not tuned to provide recommendations for better employee retention and talent management.
This study's main objective is to build a robust hybrid employee churn prediction model for the organizations. For this purpose, the factors determining employee churn were inspired by an employee record dataset from the website Kaggle (In Class Prediction Competition,2017), a data analytics platform that contributes datasets and performs research in data science.
The proposed model aims to identify the critical factors that govern employee churn and provide customized recommendations to HR managers on how to retain valuable and disgruntled employees.
This study's data is collected by interviewing Information Technology sector employees in India and this data is used for training the prediction model. The model is then validated on a real-time employee dataset of a mid-sized Fast-Moving Consumer Goods (FMCG) company for comparing the predictive power of the classifiers.
The paper is structured as follows: The literature review of the existing studies in churn prediction and the rationale for adopting the above predictors is detailed. The data collection and research methodology are then elucidated. The results and discussion are illustrated. The study is then concluded and the scope for future research is discussed. The references follow this.