Κατεύθυνση Οικονομικά, Διοικητικά και Πληροφοριακά Συστήματα Επιχειρήσεων
Library of the Faculty of Economics and of the Faculty of Business Administration

2019-11-17

2019

Giagkli Maria

Βασιλείου Ελ Ευάγγελος, Επίκουρος Καθηγητής, Τμήμα Μηχανικών Οικονομίας και Διοίκησης, Πανεπιστήμιο Αιγαίου

Διερεύνηση αιτιών εργασιακής φθοράς με χρήση μεθόδων ταξινόμησης

Greek

Investigating the reasons of employee attrition with classification methods

Employee attrition is one of the most serious issues for organizations in current global economy. Human resources can be a key factor of organizations’ competitive advantage, but at the same time a costly challenge for them. For this reason, companies want to understand better the main issues behind employee attrition phenomenon. The decision of an employee to quit may be due to a variety of reasons. However, it’s important for companies to find out the most significant of them and predict who has the highest possibility to leave, so that they proceed to some changes and make a better human resource management.
In this thesis, we chose to study statistical learning algorithms, in order to locate the main reasons of employee attrition, through data mining methods. Specifically, we analyze the algorithms of Logistic Regression, Decision Trees and Random Forest by using environment of R and R studio. The study is conducted on a dataset provided by kaggle.com, which is a data mining competition website. This data set has about 30.000 records and considers some of the employees’ characteristics, such as satisfaction level, last evaluation rating, salary and so on.
The study firstly focuses on the theoretical depiction of algorithms and then on practical implementation and analysis of our data set. After iterative parameterization and evaluation of each algorithm, we conclude to the best possible model for each of them. By comparing their performance to unknown data, we select a final model, through which we draw our conclusions. At last we provide some suggestions for further research.
This study area has a wide range of practical and useful application. If it has been resolved properly, the same method and model can be used to study other problems such as customer churn, customer relation management, youth unemployment and more.

Social, Political and Economic sciences

Statistical Learning Methods, Supervised Learning, Classification Algorithms, Logistic Regression, Decision Trees, Random Forest, Evaluation Metrics of Classification

No

0

Yes

32

96

https://pergamos.lib.uoa.gr/uoa/dl/object/2885645/file.pdf

https://pergamos.lib.uoa.gr/uoa/dl/object/2885645