基于XGBoost的自动驾驶汽车事故风险预测研究
A Study on Accident Risk Prediction For Autonomous Vehicles Based on XGBoost
朱小平 1张丽英 1刘静 1向健龙1
作者信息
- 1. 桂林电子科技大学 广西桂林市 541004
- 折叠
摘要
自动驾驶汽车风险具有复杂性和隐蔽性,不易被人为地发现和预防.为了更好地预测这些风险,利用美国加州自动驾驶事故数据集,从时间、地点、人员参与、天气等多维度提取数据,数据经过预处理从而构建自动驾驶事故数据库.然后,将XGBOOST算法与数据相结合,建立自动驾驶汽车事故风险预测分类模型.将XGBOOST算法与多种算法进行比较分析,结果表明,XGBOOST算法为较优,其训练和测试预测精度分别超过 92.27%和 97.06%,能够有效地识别出高风险和低风险的自动驾驶汽车事故情况.
Abstract
The risks of autonomous vehicles are complex and hidden,and are not easy to be discovered and prevented by human beings.In order to better predict these risks,we use the California autonomous driving accident dataset,and extract data from multiple dimensions such as time,location,personnel participation,weather,etc.The data is preprocessed to construct an autonomous driving accident database.Then,we combine the XGBOOST algorithm with the data to build an autonomous driving accident risk prediction classification model.We compare the XGBOOST algorithm with various algorithms,and the results show that the XGBOOST algorithm is superior,with training and testing prediction accuracy exceeding 92.27% and 97.06%,respectively,and can effectively identify the high-risk and low-risk situations of autonomous driving accidents.
关键词
自动驾驶汽车/XGBoost算法/风险预测Key words
Autonomous vehicles/XGBoost algorithm/Risk prediction引用本文复制引用
基金项目
桂林电子科技大学研究生教育创新计划项目(2023YCXS192)
出版年
2024
NETL
NSTL
万方数据