基于组合赋权优化的ES-ARIMA-BP神经网络交通事故预测研究  

Research on traffic accident prediction by ES-ARIMA-BP neural network based on combined empowerment optimisation

作　　者：刘尊青 单小曼 辛宁 侯金超 姚亮 钟丽华 LIU Zunqing;SHAN Xiaoman;XIN Ning;HOU Jinchao;YAO Liang;ZHONG Lihua(School of Transportation and Logistics Engineering,Xinjiang Agricultural University,Urumqi 830052,China;Traffic Police Headquarters,Public Security Department of Xinjiang Uygur Autonomous Region,Urumqi 830011,China)

机构地区：[1]新疆农业大学交通与物流工程学院,新疆乌鲁木齐830052 [2]新疆维吾尔自治区公安厅交警总队,新疆乌鲁木齐830011

出　　处：《现代电子技术》2024年第22期71-76,共6页Modern Electronics Technique

基　　金：国家自然科学基金项目(52268072);新疆农业大学校级重点学科项目(XJAUTE2022G09);新疆农业大学科研项目(2523HXKT3)。

摘　　要：为提高交通事故模型的预测精度,更好地辨识交通事故在时间维度上的规律特性,基于CRITIC法和熵权法组合赋权,构建一种ES-ARIMA-BP神经网络组合预测模型,探究新疆地区交通事故在时间维度上的月度分布规律。首先,使用指数平滑法(ES)进行预测,可减少数据间的噪声,并能捕捉时间序列数据中的季节性变动;其次,使用ARIMA模型进行预测,可捕捉数据中的线性部分、非季节性趋势和周期性波动;最后,为更好地应对数据中的复杂非线性及无周期性波动,引入BP神经网络进行预测。结果表明:构建基于组合赋权优化的ES-ARIMA-BP神经网络组合预测模型,平均绝对误差百分比(MAPE)仅为1.869%,决定系数(R^(2))高达0.982,较单一模型及单一赋权法下的组合模型预测误差率更低,拟合程度更好。组合预测模型以数据最大优化为思想基础,可有效克服单一模型的局限,同时采用组合赋权,使其能更好地适应不断变化的数据和环境,从而提高预测的准确度。In order to improve the prediction accuracy of the traffic accident model and better identify the regular characteristics of traffic accidents in the time dimension,a combined ES-ARIMA-BP neural network prediction combination model is constructed based on the combination of CRITIC and entropy weight method to explore the monthly distribution of traffic accidents in Xinjiang in time dimension.The exponential smoothing (ES) method is used for prediction,which can reduce the noise among the data and capture the seasonal variations in the time series data.Forecasting with the ARIMA model can capture the linear component,non-seasonal trends and cyclical fluctuations in the data.In order to better deal with the complex nonlinear and non-periodic fluctuations in the data,BP neural network is introduced for prediction.The results show that the mean absolute percentage error (MAPE) and coefficient of determination (R~2) of the ES-ARMI-BP neural network combination prediction model based on combinatorial weighting optimization are only 1.869% and 0.982,which is lower than the prediction error rate and better fitting degree of the combined model under the single model and the single weighting method.Based on the idea of maximum data optimization,the combinatorial forecasting model can effectively overcome the limitations of a single model,and can use combined weighting to make it better adapt to the changing data and environment,so as to improve the accuracy of prediction.

关 键 词：交通事故预测 ES-ARIMA-BP 神经网络 组合模型 预测模型 赋权优化 

分 类 号：TN911.23-34[电子电信—通信与信息系统] U491.31[电子电信—信息与通信工程]