Enhancing Employee Productivity Prediction: A CNN-GRU Approach to Workstation Activity Analysis

Employee productivity prediction is crucial for optimizing workplace efficiency and performance management. This study proposes a hybrid CNN-GRU model to analyze workstation activity patterns and predict employee productivity with high accuracy. The CNN component extracts spatial features, while the GRU component captures temporal dependencies, enabling a robust prediction framework. The model is evaluated against CNN, GRU, Random Forest, and SVM, demonstrating its superior performance. The CNN-GRU model achieves 92.5% accuracy, 91.8% precision, 90.4% recall, and an F1-score of 91.1%, outperforming CNN (86.7% accuracy) and GRU (88.3% accuracy), as well as traditional machine learning models such as Random Forest (80.9% accuracy) and SVM (78.4% accuracy). Feature importance analysis indicates that keyboard usage (30%) and mouse activity (20%) are the most influential factors in productivity prediction. Although CNN-GRU has the highest training time (220.3 seconds), its improved predictive performance justifies the computational cost. The results demonstrate that deep learning-based approaches are more effective than traditional models for productivity prediction. This research provides valuable insights into workstation activity monitoring and paves the way for real-time performance assessment and workplace optimization strategies.