고속도로 사고등급별 돌발상황 처리시간 예측모형 개발에 관한 연구

Abstract

본 연구는 고속도로에서 발생하는 비반복적 정체 중 교통사고로 인하여 발생하는 돌발상황에 대한 운영관리 부족 및 처리시간에 대한 정보제공 요구증가에 따라 이에 대응할 수 있는 돌발상황 처리시간 예측모형개발을 개발하였다. 우선 경부고속도로에서 발생한 돌발상황 이력데이터를 수집하여 교통사고등급에 따른 돌발상황 처리시간에 대한 분포를 살펴보았다. 그 결과 돌발상황 처리시간이 100분 이상인 자료는 현실적으로 흔히 발생하는 돌발상황이 아니므로 모형의 정확성을 위하여 제거한 후 데이터를 재정리하였다. 돌발상황 처리시간 예측모형을 개발하기에 앞서 종속변수인 사고처리시간을 사고등급 A, B, C등급으로 구분하였으며, 독립변수로는 교통량, 사고차량수, 사고시간대 등 총 16개 변수를 적용하여 모형을 개발하였으며, 모형의 검증은 MPB, MAD, Theil-부등계수를 이용하였다. 모형도출결과 돌발상황 처리시간에 영향을 미치는 주요변수로는 교통량, 사고등급, 중차량포함여부, 사고시간대가 도출되었으며, 모형검증결과 모형의 설명력이 있는 것으로 도출되었다. 또한 돌발상황 처리시간 예측모형에서 영향 변수로 도출된 변수들을 토대로 의사결정나무를 구축하였으며, 이때 CHAID기법을 적용하였다. 그 결과 1차적으로 사고등급 A, B등급과 C등급으로 구분되었으며, 2차적으로는 도로의 교통량으로 분리되었다. 따라서 본 연구를 통하여 도출된 돌발상황 처리시간 예측모형과 의사결정나무를 통하여 향후 고속도로 돌발상황 발생시 도로이용자들에게 보다 신속하고 실효성있는 교통정보를 제공하는데 기여할 수 있을 것으로 판단된다.; In this study, the prediction model development for incident reaction time was carried out so that we can cope with the increasing demand for information related to the reaction time of emergencies due to the traffic accidents while non-repeatable congestion is happening in expressways. First of all, the history data for incidents that occurred in Gyeongbu Expressway was collected and then the distribution for the reaction time when incidents were dealt with according to the grade for traffic accidents was reviewed. As a result, as the materials that took more than 100 minutes are not considered as the incidents, they were removed for the accuracy of the model and the data was rearranged. Before the prediction model for incident reaction time was developed, the time for dealing with accidents, the dependent variable, was classified into incident grade, A, B, and C. Then, as a dependent variable, 16 ones in total including traffic volume, the number of vehicles that accidents occurred to, and the accidents time zone were applied and thereby the model was developed. To verify the model, MPB, MAD, Theil's Inequility Coefficient were utilized. As a result of inducing the model, traffic volume, as the main variables that affect incident reaction time, traffic volume, an accident grade, possibility of including heavy vehicles, and an accident time zone were induced. The results of verifying the model showed that the model has some degree of explanatory power. In addition, the Answer Tree was constructed based on the variables induced as the influencing variable in prediction model for incident reaction time, when the CHAID Technique was applied. As a result of this, accidents were classified into grade A, grade B, and grade C first. In the secondary classification, they were divided into traffic volume. Accordingly, this study is expected to make a contribution to providing expressway users with quicker and more effective traffic information when incidents happen on expressway afterwards through the prediction model for incident reaction time and the Answer Tree.