Real-time driver stress detection is critical for road safety but faces challenges regarding the trade-off between accuracy and interpretability. While Deep Learning (DL) models achieve high accuracy, they often lack the transparency required for safety-critical vehicle applications. This paper proposes a Clustering-based Fuzzy System for Detecting Driver Stress (CFS-DDS) that balances performance with decision transparency. Using Fuzzy C-Means clustering on the AffectiveROAD dataset, CFS-DDS automatically generates membership functions from Heart Rate (HR), Electrodermal Activity (EDA), and Skin Temperature (ST) to estimate the Driver Stress Level (DSL), identifying physiological stress patterns such as “Cold Sweat”. Experimental results on 13 drivers in real-world driving conditions demonstrate that CFS-DDS achieves an average binary classification accuracy of 73.4%. This performance is comparable to existing user-independent Machine Learning (ML) baselines (e.g., Random Forest: 74.1%) while offering interpretable linguistic rules. The system achieved 88% to 93% accuracy for typical responders, effectively detecting stress events. Furthermore, this study quantitatively analyzes all drivers, clarifying the significant impact of physiological inter- and intra-subject variability on generalized models. CFS-DDS provides a transparent alternative to black-box ML/DL models for in-vehicle monitoring.