SatisStruct-Net: A Deep Learning-Based Framework for Evaluating Job Satisfaction and Performance in the Digital Transformation of Construction

The rapid adoption of digital transformation in the construction industry has introduced challenges related to job satisfaction and workforce performance. This paper presents SatisStruct-Net, a deep learning-based framework designed to evaluate and predict job satisfaction and workforce performance during digital transformation. Leveraging a hybrid architecture combining ResNet for hierarchical feature extraction and Transformers for contextual learning, the model achieves state-of-the-art results. Capsule networks ensure accurate hierarchical representation of features such as digital training exposure, task complexity, and collaboration metrics. The framework achieves a high R-squared of 0.91, RMSE of 0.42, and MAE of 0.32 for satisfaction prediction on the test set. For workforce performance classification, the model records an average precision of 0.92 and an AUC of 0.96 for top-performing departments like the Design Team. Comparative analysis shows that SatisStruct-Net outperforms baseline models such as Random Forest and XGBoost by a significant margin. Real-time dashboard integration highlights actionable insights, such as optimizing training hours and simplifying complex workflows to improve satisfaction and performance. These results validate the reliability and scalability of SatisStruct-Net as a robust solution for the construction industry.