Answers are grounded in this paper and cite their sources.
In the present era of data-driven organizational environment, the practice of Human Resource Management (HRM) has become increasingly reliant on intelligent Decision-Support Systems (DSS). This study develops a multifaceted two-pipeline model of Predictive Modelling (PM) and Sentiment Analysis (SA) to enhance workforce analytics capabilities. A publicly available HRM analytic dataset is used to train supervised classification models, including Random Forest (RF), Extreme Gradient Boosting (XGBoost), and Long Short-Term Memory (LSTM), as well as an ensemble model that integrates these classifiers. These approaches use structured data to predict employee attrition based on features such as age, job role, experience, and job satisfaction. The unstructured textual data sources, including resumes and employee reviews, are handled using state-of-the-art Natural Language Processing (NLP) such as tokenization, Term Frequency-Inverse Document Frequency (TF-IDF), and Bidirectional Encoder Representations as Transformers (BERT)-based embeddings. The new Mathematically Modified Robustly Optimized BERT Pretraining (MM-RoBERTa) is proposed for extracting the PM and SA. All the models are evaluated using k-fold Cross-Validation (CV) and standard evaluation measures, namely Accuracy, F1-score, Area Under the Receiver Operating Characteristic Curve (AUC), and Mean Absolute Error (MAE). The ensemble model achieves a predictive accuracy of 91.3%, and MM-RoBERTa outperforms existing SA with an accuracy of 93.1 %. The combination of predictive and affective insights is of practical use in fine-tuning talent retention, empowering HRM professionals to make informed decisions based on objective performance indicators and subjective emotional states.
No takes yet. Share an insight, caveat, or question.
Tripathi et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: