Wang Tao, Chu Jie, Yao Zhi-Ying, Guo Xiao-Lei, Ma Yu-Bin, Lu Zi-Long, Jia Cun-Xian
Phase 1 Clinical Trial Center, Deyang People's Hospital, Deyang, Sichuan, China.
Department of Epidemiology, School of Public Health, Cheeloo College of Medicine, Shandong University, Jinan, China.
Sci Rep. 2025 Apr 26;15(1):14620. doi: 10.1038/s41598-025-99433-7.
To predict and analyze road traffic crash-related mortality rates across different groups in Shandong Province using the GM (1,1) model and the GM-BP joint model. We also sought to establish the optimal model and provide a theoretical basis for road traffic crash prevention and control. Herein, road traffic fatality-related data between 2012 and 2022 in Shandong Province were collected using the ICD-10 codes in the Population Death Information Registration Management System of the Chinese Center for Disease Control and Prevention. After cleaning and collating the data, the Grey Modeling Software was used to construct the GM (1,1) model and the SPSSPRO software was used to train the BP neural network model to build the GM-BP joint model based on relevant information from the GM (1,1) model. The two predictive models were evaluated using Mean Squared Error (MSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and Root Mean Squared Error (RMSE). Between 2012 and 2022, there were 176,129 victims of road traffic crash-related fatalities in Shandong Province, with an average age of 52.65 ± 17.94 years. The gender ratio was around 2.77:1, and the average standardized mortality rate for the total population was 18.59/100,000 persons. According to the model evaluation results, compared to the GM-BP joint model, the GM (1,1) model had smaller MSE, MAE, MAPE, and RMSE values for the total population and motorized drivers but larger values for pedestrians, non-motorized drivers, and passengers. In cases of slightly changing data, the GM-BP joint model can effectively leverage the advantages of the GM (1,1) model and the BP neural network model, improving the prediction accuracy and reliability. Our findings could provide critical supportive data and decision-making references for road traffic management authorities, facilitating the development of prevention and control programs for road traffic safety tailored for specific groups of road users (e.g., non-motorized drivers).
使用GM(1,1)模型和GM-BP联合模型预测和分析山东省不同群体的道路交通事故相关死亡率。我们还试图建立最优模型，并为道路交通事故的预防和控制提供理论依据。在此，利用中国疾病预防控制中心人口死亡信息登记管理系统中的ICD-10编码收集了山东省2012年至2022年期间与道路交通事故死亡相关的数据。在对数据进行清理和整理后，使用灰色建模软件构建GM(1,1)模型，并使用SPSSPRO软件训练BP神经网络模型，基于GM(1,1)模型的相关信息构建GM-BP联合模型。使用均方误差(MSE)、平均绝对误差(MAE)、平均绝对百分比误差(MAPE)和均方根误差(RMSE)对这两种预测模型进行评估。2012年至2022年期间，山东省共有176,129名道路交通事故相关死亡受害者，平均年龄为52.65±17.94岁。性别比约为2.77:1，总人口的平均标准化死亡率为18.59/10万。根据模型评估结果，与GM-BP联合模型相比，GM(1,1)模型在总人口和机动车驾驶员方面的MSE、MAE、MAPE和RMSE值较小，但在行人、非机动车驾驶员和乘客方面的值较大。在数据略有变化的情况下，GM-BP联合模型可以有效利用GM(1,1)模型和BP神经网络模型的优势，提高预测的准确性和可靠性。我们的研究结果可以为道路交通管理部门提供关键的支持数据和决策参考，促进针对特定道路使用者群体(如非机动车驾驶员)制定道路交通安全预防和控制方案。
