A Novel Approach to Driver Negligence Detection: <scp>EAXB</scp>‐<scp>EVS</scp> Algorithm With <scp>IoT</scp> Integration

ABSTRACT In recent days, road accidents have become a major issue, caused by negligence of the driver such as drowsiness, alcohol consumption during driving, gas leakage, tiredness, and traffic law violations. Timely decisions and accurate driver negligence detection are mandatory for avoiding road accidents and fatalities. However, the earlier research encountered various obstacles such as more energy consumption, poor detection performance, and required high computational time. This research proposes a novel Internet of Things‐based Ensemble‐voting Adaptive Extreme Boost‐based Energy Valley Search algorithm within the Vehicular Ad‐Hoc Networks environment for early and accurate detection of driver negligence. The proposed Ensemble‐voting Adaptive Extreme Boost‐based Energy Valley Search framework utilizes an Eye Aspect Ratio and Mouth Aspect Ratio analysis to identify the driver's facial contours like mouth and pupils by analyzing the imaging data. Further, the Ensemble‐voting Adaptive Extreme Boost model is employed for effective prediction of driver negligence that combines the weak learner models using ensemble voting and adapts dynamically for identifying the driver's state of carelessness including intoxication, gas leakage, and drowsiness. The hyperparameter of the Ensemble‐voting Adaptive Extreme Boost framework is fine‐tuned by applying Energy Valley Search, which enhances the accuracy and efficiency of the system while minimizing the computational overhead and energy consumption. The effectiveness of the proposed model is validated using some evaluation measures on three datasets namely, the Driver Inattention Detection Dataset, the India Road Accident Dataset, and the Driver Drowsiness Dataset. The simulation outcomes indicate that the Ensemble‐voting Adaptive Extreme Boost‐based Energy Valley Search framework attained a higher accuracy of 98.87%, less execution time of 1.03 s, and less energy usage of 5%, which makes the proposed system highly efficient for the real‐time vehicular network application.