A recommendation approach for employee retention by using a new feature selection strategy

Date

2023-9

Author

Taşkıran, Nagihan

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

194
views

164
downloads

Cite This

Nowadays, keeping experienced and talented employees is one of the major concern of the organizations. Extra costs related with hiring and training of new employee, loss of know-how and customer satisfaction makes employee retention more crucial for organizations. Therefore, Human Resources (HR) Department of companies try many strategies to keep valuable employees in the long period of time. However, a strategy appealing to a group of people and ignoring personal priorities may not be successful or appropriate for every employee in that group. At this point, improving effective recommendation strategies to keep the employee gains importance for HR departments. There are many studies in the literature, which focus on the attributes that lead the employees to leave, but fewer studies take into account personal reasons. This study focuses on defining the most effective attributes that lead each precious employee to churn by providing an improved feature selection method and recommendation algorithm. To define precious employees for the company Fuzzy C-Means clustering algorithm is applied. An enhanced feature selection method that uses different ensemble learning algorithms together with different feature selection approaches, including Forward and Backward Feature Selection is proposed. The method considers a certain number of most voted attributes among all different combinations of classification and feature selection methodologies. After choosing these attributes, several classification algorithms are applied including Random Forest (RF), Extreme Gradient Boosting (XGB), and Gradient Boosting (GB) for prediction of potential churner employees. In addition, a different recommendation approach is released based on the similarity between possible churner and non-churner employees. The main idea behind this approach is finding different attributes that can affect churn intention of an employee by comparing a non-churner employee and a potential churner employee who have similar features uncontrolled by HR. Finally, n attributes are offered to the attention of HR department for each possible churner employee which are categorized as medium and experienced.

Subject Keywords

Recommender systems, Extreme gradient boosting, Feature selection, Hyper-parameter optimization

URI

https://hdl.handle.net/11511/105412

Collections

Graduate School of Natural and Applied Sciences, Thesis