Improved Machine Learning Prediction Framework for Employees with a Focus on Function Selection

Companies are constantly looking for methods to keep their workers with them to minimize further recruitment and training expenses. Predicting whether a specific staff member can depart helps the business to take preventative measures. Unlike physical systems, a scientific and analytical formula cannot explain human resource issues. Machine learning methods are thus the ideal instruments for this purpose. This chapter offers a three-stage paradigm for the prevention of attrition (up to processing, processing, post-processing). IBM HR dataset for the case study is selected. As there are many functions in the dataset, the selection technique for the “maximum-out” feature is suggested for the extent of decreasing the to-processing phase. In the planning retrogression model, the coefficient of each variable indicates the significance of the feature in attrition predictions. The findings indicate an improvement in the F1 score using the “maximum-out” feature selection technique. Finally, through learning the model for many bootstrap datasets, the validity of data is verified. The average deviation of the parameters is then evaluated to verify the confidence and stability of the model parameters. The modest average deviation of the data shows the model is stable and generalized.