**5. Conclusions**

This study presented the findings on the levels of driving stress depending on the tra ffic conditions and road types, using the EDA signals. The study confirmed that EDA is a significant indicator of the psychological stress and an e ffective tool to determine the stress levels in real driving conditions. Advantages of the developed method are that both developed models have an ability to predict stress levels under actual driving conditions with over 80% accuracy. The developed models can also be used to classify driving stress in di fferent tra ffic situations and di fferent road types. The definition of a tra ffic jam state in a driving situation is one of the important contributions of this study. One of the characteristics of the models is the ability to classify the stress level without considering the rest state of the driver. This makes the models universal for use in a variety of situations without the preliminary intervention. The proposed model and the experimental procedure are expected to be easily reconfigured by other researchers. One of the disadvantages of the proposed model is that the tra ffic congestion concept needs to be defined and improved more specifically based on tra ffic flow theory. To make a better model in the future, the following points should be considered. First, researchers can collect more road information, such as lane-keeping status, as well as bio-signals, such as ECG. In addition, various kinds of machine-learning techniques can be applied instead of logistic regression. Finally, researchers can incorporate other factors, such as seasonal factors, in-vehicle temperature, and a driver's attention into the model as key features.

Obtained results can be used for practical application. Examples of such uses are sensors for monitoring of the autonomic nervous system, smartphone/computer applications and wearable wireless devices with EDA sensors.

**Author Contributions:** J.P. proposed the main idea and conducted the experiment; O.V.B. finished the draft manuscript and drew the figures and tables; J.P. and J.K. analyzed the data; H.K.K. revised and finished the manuscript.

**Funding:** This work was supported by the Incheon National University Research Grant in 2017 (Grant No.: 20170467).

**Conflicts of Interest:** The authors declare no conflict of interest.
