The invention relates to a short-time traffic flow prediction method and system, an electronic device and a storage medium, and the method comprises the steps: obtaining multi-mode road network sourcedata, carrying out the preprocessing of the data through employing a road segment and time as the main dimensions, and generating basic data with the unified caliber and road segment matching; respectively carrying out feature extraction on the road network spatio-temporal data, environment data and video image data in the basic data to obtain spatio-temporal dependence features, environment features and image features; performing feature fusion on the obtained features to obtain fused features; filtering the fusion features through an attention mechanism model, and capturing key effective features in the fusion features; and finally, inputting the effective features into a regression model for prediction to obtain a prediction value. Through the method and system, the problems of low road network multi-modal data utilization rate and low road condition multi-condition prediction accuracy are solved, and the model prediction accuracy and efficiency are improved.
本申请涉及短时交通流量预测的方法、系统、电子装置和存储介质,其中,该短时交通流量预测方法包括:获取多模态路网源数据,以路段和时间为主维度,对数据进行预处理,生成口径统一和路段匹配的基础数据;分别对基础数据中的路网时空数据、环境数据和视频图像数据进行特征提取,获取时空依赖特征、环境特征以及图像特征;对获取的特征进行特征融合,获得融合特征;通过注意力机制模型对融合特征进行过滤,捕捉融合特征中的关键有效特征;最终将有效特征输入回归模型中进行预测,得到预测值。通过本申请,解决了路网多模态数据利用率低和路况多情况预测准确率低的问题,提高了模型预测准确性和高效性。
Short-term traffic flow prediction method and system, electronic device and storage medium
短时交通流量预测的方法、系统、电子装置和存储介质
2021-03-02
Patent
Electronic Resource
Chinese
Short-term traffic flow prediction method and device, electronic equipment and storage medium
European Patent Office | 2024
|Short-term traffic flow prediction method and system, storage medium, equipment and terminal
European Patent Office | 2024
|European Patent Office | 2022
|