Predictive Analytics for Employee Retention UsingWorkday HCM Data

Abstract

The retention of employees is an important aspect
of organizational stability and performance because high rates
of employee attrition mean that the organization incurs
financial losses, decreased productivity, and institutional
knowledge loss. Predictive analytics is a proactive way of
comprehending and predicting employee turnover in using
demographic, behavioral, and work-related information. A
practical model for predicting employee retention is the goal of
this research, which applies machine learning techniques to the
1,470-item Employee Attrition and Performance dataset from
IBM HR Analytics. The dataset contains 35 different qualities.
Data preprocessing such as the processing of missing values,
outlier processing, label encoding, normalization, and feature
selection were done comprehensively to guarantee data quality.
The problem of the imbalance of classes was addressed by
creating balanced training data with the SMOTE technique.
The postulated model employs the Extra Trees Classifier (ETC)
that is an ensemble learning algorithm that constructs
numerous randomized decision trees to improve the predictive
capability and strength. The ETC model was doing
exceptionally well with an accuracy (ACC) of 99.1%, precision
(PRE) of 98.6%, recall (REC) of 99.8% and F1-score(F1) of
99.2%. These results affirm the applicability and relevance of
the model in employee retention prediction as it provides the
organization with practical data of how to optimize HR
practices and support businesses to make informed judgments
concerning workforce management.