Theodoros Garefalakis, “Identification of driver’s risky behavior level and duration with machine learning techniques”, Diploma Thesis, NTUA, School of Civil Engineering, Athens, March 2022
ID
ad125
Presentation
Full Text
Publications
The objective of this Thesis is the identification of driver’s risky behavior level and duration with machine learning techniques. For this purpose, useful data related to driving behavior were collected through a driving simulator experiment. Based on the processing and analysis of the data, three levels of risk were defined. In the first part of the analysis, four machine learning algorithms were developed to classify driver behavior into one of three risk levels, with the 'Random Forests' algorithm scoring the highest performance. In the context of investigating the influence of driving factors to identify driving behavior, the
distance traveled, speed and speed limit emerged as the most important. In the second part of the analysis, the effect of driving characteristics on driving duration at different stages was examined. To achieve the above goal, three regression algorithms were developed to predict driving duration at each safety level. The effect of different variables on the forecasting process was determined based on the performance of the models and their statistical significance. The results showed that the maximum speed was the most important variable, which negatively affects the driving duration at each safety level.