Predicting Employee Attrition using Machine Learning
Прогнозирование увольнения сотрудников с использованием машинного обучения
2018-11-01
SCID: 54.1/gfa38hxd
Discuss with AI
ADASYNF1-scoreIBM Watson synthetic dataK-nearest neighbour (KNN)class imbalanceemployee attritionfeature selectionmachine learningrandom forestsupport vector machine (SVM)undersampling
Figures from the paper
Abstract (AI)
The growing interest in machine learning among business leaders and decision makers demands that researchers explore its use within business organisations. One of the major issues facing business leaders within companies is the loss of talented employees. This research studies employee attrition using machine learning models. Using a synthetic data created by IBM Watson, three main experiments were conducted to predict employee attrition. The first experiment involved training the original class-imbalanced dataset with the following machine learning models: support vector machine (SVM) with several kernel functions, random forest and K-nearest neighbour (KNN). The second experiment focused on using adaptive synthetic (ADASYN) approach to overcome class imbalance, then retraining on the new dataset using the abovementioned machine learning models. The third experiment involved using manual undersampling of the data to balance between classes. As a result, training an ADASYN-balanced dataset with KNN (K = 3) achieved the highest performance, with 0.93 F1-score. Finally, by using feature selection and random forest, F1-score of 0.909 was achieved using 12 features out of a total of 29 features.
Key Findings
1
Feature selection with random forest reduced features from 29 to 12 and achieved an F1-score of 0.909
2
Manual undersampling was evaluated as an alternative class-balancing approach (results not exceeding ADASYN+KNN)
3
Three experiments were conducted to predict employee attrition using SVM (various kernels), random forest, and KNN on IBM Watson synthetic data
4
Training on the original imbalanced dataset with SVM, random forest, and KNN was performed as a baseline
5
Using ADASYN to address class imbalance and training KNN with K=3 achieved the highest performance with a 0.93 F1-score
Research Object
Employee attrition in business organizations (synthetic IBM Watson employee dataset)
Research Subject
Predictive modeling of employee attrition using machine learning approaches, including class-imbalance handling (ADASYN, undersampling), model comparison (SVM, random forest, KNN), and feature selection to optimize F1-score
Publication Details
Publication Date
2018-11-01
Journal
Publisher
ISSN
Cited by
143
Access Type
Author Information
Download PDF
Subscribe to digest