A High-Precision Driving Behavior Classification Framework using Convolutional Neural Networks based on Vehicle Telematics Data

The increase in road travel has a significant impact on road safety, fuel consumption, and environmental pollution. Continuous efforts have been made to increase road safety and reduce fuel consumption and pollution. One of the factors that is of great importance in this regard is driver behavior. In this study, an advanced artificial intelligence model using convolutional neural networks is used to accurately classify driving behaviors under speed threshold criteria. In this modeling, a dataset with 370 data points is used with five inputs. The input data for modeling and predicting driver behaviors include fuel flow rate/hour, engine speed, speed, acceleration, and grade. Four classes (labels) are considered: normal (class 1), moderate (class 2), aggressive (class 3), and dangerous (class 4). The obtained results indicate that the convolutional neural network model achieved very high classification performance during the modeling process. In addition to classification accuracy, the proposed framework contributes by establishing a structured behavior classification scheme based on telematics data and speed-threshold criteria, which can support intelligent safety monitoring, eco-driving assessment, and transportation management applications. Finally, the most important factors that affect driver behavior classification are the speed of cars and engine speed.