다변화된 시간 집계단위 및 동적 교통류 특성을 고려한 안전성능함수 개발 연구

초록

This study aims to develop a safety performance function (SPF) that reflects dynamic traffic characteristics based on highway traffic crash data, and to improve the prediction performance by complementing the limitations of the existing macroscopic traffic aggregation unit. The VDS data was aggregated into annual average daily traffic (AADT), annual hourly average traffic (AHT), and annual average 15-minute traffic (AMT), and negative binomial regression (NB), random forest (RF), XGBoost, and LGBM models were developed for each dataset to compare the prediction performance. The results showed that the XGB model was the best model for daily traffic data, and the LGBM model showed good prediction performance for hourly and 15-min data. This study improved the crash prediction model by setting a detailed unit, which is different from the existing macroscopic aggregation unit.

키워드