<![CDATA[Microsleep presents a critical safety challenge for Indonesian drivers, particularly affecting long-distance transportation where existing detection methods remain costly and impractical for widespread deployment. This study introduces a novel application of Random Forest algorithm specifically tailored to Indonesian driving contexts, utilizing locally-sourced accident data combined with driver behavioral surveys to predict microsleep likelihood. Unlike previous studies that relied primarily on physiological monitoring or international datasets, this research leverages accessible vehicle and environmental variables including driving duration, road conditions, weather patterns, and work schedules from National Transportation Safety Committee (KNKT) records spanning 2013-2023. The Random Forest model, configured with 100 trees and maximum depth of 10, demonstrated 87.50% overall accuracy with perfect recall (1.00) for microsleep detection when validated using stratified k-fold cross-validation. This study uniquely contributes to the field by demonstrating that context-specific environmental and behavioral factors can effectively predict microsleep incidents without expensive physiological monitoring, offering a practical foundation for developing cost-effective vehicle safety systems tailored to Indonesian road conditions and driving patterns. The findings provide actionable insights for transportation policy development and establish a framework for implementing affordable microsleep detection in developing countries with similar traffic characteristics.]]>
Copyrights © 2025
