The invention belongs to the technical field of network models, and particularly relates to a road traffic accident two-stage generation enhanced network model considering unbalanced data, comprising the following steps: collecting data, and preprocessing the data; sMOTE is improved based on simulated annealing to solve the problems of parameter fixation and noise influence in a neighbor interpolation process; the generative adversarial network is embedded with an auto-encoder component; performing classification prediction by using an artificial neural network; and performing model interpretation by using the SHAP. According to the method, the enhanced network model (AGSMF) is generated by combining the two stages of SA-SMOTE and AE-GAN, and each stage and each model component aim to give full play to the advantages to the greatest extent. According to the method, the model result is explained through the SHAP method, the importance of key factors in traffic accident severity prediction is revealed, and deep insight and scientific basis are provided for formulating an effective traffic accident prevention strategy.
本发明属于网络模型技术领域,具体涉及一种考虑不平衡数据的道路交通事故双阶段生成增强网络模型,包括下列步骤:收集数据,并对数据进行预处理;基于模拟退火改进SMOTE解决近邻插值过程中参数固定和噪声影响的问题;嵌有自编码器组件的生成对抗网络;利用人工神经网络进行分类预测;使用SHAP进行模型解释。本发明结合SA‑SMOTE和AE‑GAN的双阶段生成增强网络模型(AGSMF),每个阶段以及每个模型组件都旨在最大限度地发挥其优势。本发明通过SHAP方法对模型结果进行解释,不仅揭示了交通事故严重程度预测中关键因素的重要性,还为制定有效的交通事故预防策略提供了深入洞察和科学依据。
Road traffic accident two-stage generation enhanced network model considering unbalanced data
考虑不平衡数据的道路交通事故双阶段生成增强网络模型
02.08.2024
Patent
Elektronische Ressource
Chinesisch
British Library Online Contents | 1995
|DETECTING TRAFFIC AND ACCIDENT DATA AT ROAD JUNCTIONS
Europäisches Patentamt | 2015
|