Abstract

The Employee attrition is a significant issue and concern for numerous organizations and management teams globally. Attrition has negative effect on the company productivity. Due to its effect, it has attracted many research works. This study aimed at developing employee attrition predictive model. The study adopted IBR HR dataset obtained from kaggle.com repository; it consists of 1470 instances and 34 features. PCA was used for feature reduction. The predictive model was created utilizing RF, BILSTM, SVM and LSTM. The result of the attrition model developed shows the accuracy score of RF, BILSTM, SVM and LSTM to be 86%, 87%, 88% and 87% respectively. Also, the Precision values of 0.87, 0.87, 0.91 and 0.90 for RF, BILSTM, SVM and LSTM respectively. The model also demonstrated high recall value and F1 score. The confusion matrix analysis of the model shows that 221 samples were classified as TP, 4 samples were identified as TN, 34 samples were classified as FP and 35 were identified as FN. The study concluded that SVM perform better than BILSTM RF and LSTM in terms of the validating metric used. The future work can be done using another dataset and other deep learning methods can be employed. The future study can be treated as regression problem to see the performance. The study recommended that the model can be used to assist in decision making as regards attrition issues.