Year: 2024
Type: article
Abstract: Modern transportation faces a crucial challenge in ensuring road safety by addressing driving behavior concerns. This paper introduces an innovative deep learning model derived from a cellphone-collected Driving Behavior dataset, focusing on detecting and classifying aggressive driving. Using a cohort-based dataset, a hyper-deep learning model categorizes drivers into normal, slow, and aggressive groups. The system employs pre-processing methods and two methodologies, directly inputting data and incorporating feature selection. The hyper-CNN-Dense model, used for training, shows promising results. Feature selection techniques like SVD6 and MI6 achieve optimal outcomes, with a 100% accuracy rate in detecting aggressive driving. Notably, SVD6 boasts a short processing time of just 43 seconds. This research successfully identifies aggressive driving behavior with impeccable accuracy and in a remarkably short timeframe.
Authors: Noor Walid Khalid, Wisam Dawood Abdullah
Institution: University of Tikrit
Language: English
Topic: Traffic and Road Safety
Subfield: Safety, Risk, Reliability and Quality
Field: Engineering
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W4403209729