혼합효과 기계학습모형을 활용한 교통사고 예측 및 요인분석 연구
Modeling Crash Accident Based on Mixed Effect Machine Learning Considering Unobserved Heterogeneity
2
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
교통사고는 매년 수많은 인명피해, 경제적 피해를 야기해 이를 사전에 예측·예방할 필요가 있으며 사고를 사전에 예방하기 위해 수많은 연구가 진행되었다. 지난 연구의 성과로 교통사고의 예측 모형의 성능은 크게 향상되었으나 교통사고의 인적·도로 구조적·사회경제적 요인 등 다양한 요인이 복합적으로 작용해 발생하는 관측되지 않는 이질성의 해석, 통계적 모형의 사전가정의 한계 그리고 범주형 변수의 모델링은 여전히 교통사고를 모델링하는데 큰 걸림돌이 되고 있다. 따라서, 본 연구는 비선형성·범주형 변수의 해석·이질성의 통제를 위해 앙상블 기계학습모형 중 높은 성능을 보이는 XGBoost와 혼합효과모형을 결합한 알고리즘으로 교통사고를 분석하였다. 그리고, 기계학습모형의 Black box 문제점을 극복하고자 대리모형(SHAP)을 활용해 사고요인을 전역적·지역적 관점에서 분석하였으며 지역별 사고대응방향성을 시사하였다.
Abstract
Traffic accidents cause numerous casualties and economic losses annually, emphasizing the need for their prediction and prevention. Extensive research has been conducted to anticipate accidents preemptively. While previous studies have significantly enhanced the performance of predictive models for traffic accidents, challenges persist due to the complex interplay of various factors such as human, roadway, and socio-economic elements, leading to unobserved heterogeneity and limitations in linear assumptions of statistical models. Additionally, categorical variable modeling remains a significant obstacle in accident modeling. Therefore, this study aims to address the unobserved heterogeneity by employing a Mixed Effects Boosting model, which combines mixed effects modeling to alleviate linearity assumptions and a Boosting algorithm for categorical variable modeling. This approach aids in predicting traffic accidents. Moreover, to overcome the "Black box" issue inherent in machine learning models, surrogate models (SHAP) are utilized to analyze accident factors from both global and local perspectives, providing insights into regional accident response strategies.
참고문헌 (18)
- Anastasopoulos, P. Ch, "A study of factors affecting highway accident rates using the random-parameters tobit model", Accident Analysis & Prevention, 45, pp.628~633 (2011)[DOI]
- Barua, S., "Multivariate random parameters collision count data models with spatial heterogeneity", Analytic Methods in Accident Research, 9, pp.1~15 (2016)
- Gu, Y., "Predicting intersection crash frequency using connected vehicle data : A framework for geographical random forest", Accident Analysis & Prevention, 179, pp.106880~ (2023)
- Hajjem, A., "Mixedeffects random forest for clustered data", Journal of Statistical Computation and Simulation, 84(6), pp.1313~1328 (2014)
- Hastie, T., "The elements of statistical learning: Data mining, inference, and prediction" (2009)
- Lord, D., "The statistical analysis of crash-frequency data : A review and assessment of methodological alternatives", Transportation Research Part A : Policy and Practice, 44(5), pp.291~305 (2010)
- Mannering, F., "Big Data, Traditional Data and the Tradeoffs Between Prediction and Causality in Highway-Safety Analysis", Analytic Methods in Accident Research, 25 (2020)
- Mannering, F. L., "Unobserved heterogeneity and the statistical analysis of highway accident data", Analytic Methods in Accident Research, 11, pp.1~16 (2016)
- Pu, Z., "Evaluating the Nonlinear Correlation between Vertical Curve Features and Crash Frequency on Highways Using Random Forests", Journal of Transportation Engineering, Part A: Systems, 146(10) (2020)
- Sigrist, F., "Latent Gaussian Model Boosting", IEEE Transactions on Pattern Analysis and Machine Intelligence, 45(2), pp.1894~1905 (2023)
- Tang, J., "Spatial heterogeneity analysis of macro-level crashes using geographically weighted Poisson quantile regression", Accident Analysis & Prevention, 148, pp.105833~ (2020)
- Washington, S., "Statistical and Econometric Methods for Transportation Data Analysis", pp.2003, Chapman and Hall
- Global status report on road safety, "WHO", WHO, ~ (2018)
- Yan, M., "Traffic Accident Severity Prediction Based on Random Forest", Sustainability, 14(3), pp.1729~ (2022)
- Yan, Y., "Crash prediction based on random effect negative binomial model considering data heterogeneity", Physica A : Statistical Mechanics and its Applications, 547, pp.123858~ (2020)
- Yu, H., "Inference in Multiscale Geographically Weighted Regression", Geographical Analysis, 52(1) (2020)