An analytical approach to predicting employee attrition: Ensemble learning method

Mustafa Yurtsever

Makale detayları

An analytical approach to predicting employee attrition: Ensemble learning method

Yazarlar:

Mustafa Yurtsever

Yayıncı:

Afyon Kocatepe University

Ana Başlıklar:

Yapay Zeka (Diğer)

Dergi:

Afyon Kocatepe Üniversitesi İktisadi ve İdari Bilimler Fakültesi Dergisi Vol.26 Issue Özel Sayı, pp.150-160

Belge Tipi:

Makale

Yayın Tarihi:

10/2024

Belge Dili:

Türkçe

Document ID (Org No.):

2001310105

pISSN:

1302-1966

Anahtar kelimeler:

Özet:

 Companies seek various measures to prevent the departure of professional employees and thereby reduce recruitment and training costs. Predicting whether a specific employee will leave or not enables the company to take the necessary steps to minimize such losses. Therefore, predicting the likelihood of employee attrition in advance can assist employers in making strategic decisions. Employee attrition plays a crucial role in understanding employees intentions to leave their jobs. Ensemble learning models have the potential to achieve more accurate and reliable results by combining various perspectives of multiple algorithms. In this study, a dataset indicating the attrition status of employees, prepared by IBM Watson Analytics and consisting of 1,470 records, was utilized. For this purpose, the performances were evaluated using Random Forest, Support Vector Machines, Multi-Layer Perceptron model, and two different ensemble learning models. To overcome class imbalance, the adaptive synthetic data generation ADASYN approach was used. As a result, a stacked ensemble learning model trained on a balanced dataset created using ADASYN achieved an accuracy rate of 0.96.

 Companies seek various measures to prevent the departure of professional employees and thereby reduce recruitment and training costs. Predicting whether a specific employee will leave or not enables the company to take the necessary steps to minimize such losses. Therefore, predicting the likelihood of employee attrition in advance can assist employers in making strategic decisions. Employee attrition plays a crucial role in understanding employees intentions to leave their jobs. Ensemble learning models have the potential to achieve more accurate and reliable results by combining various perspectives of multiple algorithms. In this study, a dataset indicating the attrition status of employees, prepared by IBM Watson Analytics and consisting of 1,470 records, was utilized. For this purpose, the performances were evaluated using Random Forest, Support Vector Machines, Multi-Layer Perceptron model, and two different ensemble learning models. To overcome class imbalance, the adaptive synthetic data generation ADASYN approach was used. As a result, a stacked ensemble learning model trained on a balanced dataset created using ADASYN achieved an accuracy rate of 0.96.

 Companies seek various measures to prevent the departure of professional employees and thereby reduce recruitment and training costs. Predicting whether a specific employee will leave or not enables the company to take the necessary steps to minimize such losses. Therefore, predicting the likelihood of employee attrition in advance can assist employers in making strategic decisions. Employee attrition plays a crucial role in understanding employees intentions to leave their jobs. Ensemble learning models have the potential to achieve more accurate and reliable results by combining various perspectives of multiple algorithms. In this study, a dataset indicating the attrition status of employees, prepared by IBM Watson Analytics and consisting of 1,470 records, was utilized. For this purpose, the performances were evaluated using Random Forest, Support Vector Machines, Multi-Layer Perceptron model, and two different ensemble learning models. To overcome class imbalance, the adaptive synthetic data generation ADASYN approach was used. As a result, a stacked ensemble learning model trained on a balanced dataset created using ADASYN achieved an accuracy rate of 0.96.

 Şirketler, profesyonel çalışanlarının ayrılmasını engelleyerek işe alım ve eğitim maliyetlerini azaltmak için çeşitli önlemler ararlar. Belirli bir çalışanın ayrılıp ayrılmayacağını önceden tahmin etmek, şirketin bu tür kayıpları minimize etmek için gereken adımları atmasını sağlar. Bu nedenle, çalışanların ayrılma olasılığını önceden tahmin etmek, işverenlere stratejik kararlar almalarında yardımcı olabilir. Çalışan yıpranması bu noktada çalışanların işten ayrılma niyetlerini anlama konusunda belirleyici olabilmektedir. Topluluk öğrenme modelleri, birden fazla algoritmanın çeşitli görüşlerini birleştirerek daha doğru ve güvenilir sonuçlar elde etme potansiyeline sahiptir. Bu çalışmada, 1.470 kayıttan oluşan IBM şirketi Watson Analytics tarafından hazırlanan çalışanların yıpranma durumunu gösteren veri seti kullanılmıştır. Bu amaçla, Rastgele Orman, Destek Vektör Makineleri, Çok Katmanlı Algılayıcı modeli ve iki farklı topluluk öğrenme modeli kullanılarak performansları değerlendirilmiştir. Sınıf dengesizliğini aşmak için adaptif sentetik veri üretimi ADASYN yaklaşımı kullanılmıştır. Sonuç olarak, ADASYN kullanılarak oluşturulan dengeli bir veri kümesi ile eğitilen yığın topluluk öğrenme modeli ile 0.96 doğruluk oranı elde etmiştir.

Makale detayları

An analytical approach to predicting employee attrition: Ensemble learning method

Yazarlar:

Mustafa Yurtsever

Yayıncı:

Afyon Kocatepe University

Ana Başlıklar:

Yapay Zeka (Diğer)

Dergi:

Afyon Kocatepe Üniversitesi İktisadi ve İdari Bilimler Fakültesi Dergisi Vol.26 Issue Özel Sayı, pp.150-160

Belge Tipi:

Makale

Yayın Tarihi:

10/2024

Belge Dili:

Türkçe

Document ID (Org No.):

2001310105

pISSN:

1302-1966

Employee Attrition

Prediction

Ensemble Learning MethodsÇalışan Yıpranması

Tahminleme

Topluluk Öğrenmesi Yöntemleri

Sayı

Makale detayları

An analytical approach to predicting employee attrition: Ensemble learning method

Yazarlar: Mustafa Yurtsever

Yayıncı: Afyon Kocatepe University

Ana Başlıklar: Yapay Zeka (Diğer)

Dergi: Afyon Kocatepe Üniversitesi İktisadi ve İdari Bilimler Fakültesi Dergisi Vol.26 Issue Özel Sayı, pp.150-160

Belge Tipi: Makale

Yayın Tarihi: 10/2024

Belge Dili: Türkçe

Document ID (Org No.): 2001310105

pISSN: 1302-1966

Employee Attrition

Prediction

Ensemble Learning MethodsÇalışan Yıpranması

Tahminleme

Topluluk Öğrenmesi Yöntemleri

Sayı

Yazarlar:

Mustafa Yurtsever

Yayıncı:

Afyon Kocatepe University

Ana Başlıklar:

Yapay Zeka (Diğer)

Dergi:

Afyon Kocatepe Üniversitesi İktisadi ve İdari Bilimler Fakültesi Dergisi Vol.26 Issue Özel Sayı, pp.150-160

Belge Tipi:

Makale

Yayın Tarihi:

10/2024

Belge Dili:

Türkçe

Document ID (Org No.):

2001310105

pISSN:

1302-1966