From Data to Decisions: A Novel Bert-Based Framework for Enhancing Managerial Strategies through Employee Sentiment Analysis

Abstract

Background: In today’s data-driven business landscape, employee feedback serves as a critical resource for informed managerial decision-making. This paper investigates sentiment analysis of employee textual data using both traditional and modern machine learning approaches. Baseline models like Logistic Regression and Support Vector Machine (SVM) were first utilized with conventional textual features. However, these models demonstrated limitations in handling nuanced sentiments and contextual dependencies. To overcome these challenges, a more sophisticated deep learning-based approach is suggested. The study utilizes a real-world dataset sourced from Indeed, comprising a diverse set of employee reviews. Empirical evaluations reveal that the modern approach significantly outperforms traditional techniques in terms of classification accuracy. The resulting sentiment insights offer precise, actionable guidance that can help managers fine-tune their strategies, address workplace concerns proactively, and foster continuous organizational improvement.

Objective: This research aims to analyze employee feedback using traditional machine learning and deep learning methods. It aims to create a BERT-CNN hybrid model that effectively captures contextual sentiment, offering valuable insights for informed decision-making within organizations.

Methodology: In this study, the transformer model DistilBERT is utilized alongside traditional classifiers like Support Vector Machine (SVM) and Naive Bayes to conduct sentiment analysis on employee feedback data. Hyperparameter tuning is carried out using Grid Search for all models to ensure optimal performance.

Conclusion: The proposed BERT-CNN The hybrid model demonstrated superior performance, achieving an accuracy of 95%, significantly outperforming traditional models like SVM and Naive Bayes. This framework effectively captures contextual sentiments in employee feedback, providing actionable insights for data-driven decision-making in organizational settings.