摘要
时空数据因具有时空关联性与动态演化性,导致特征挖掘难度大。目前单一维度分析方法难以全面捕捉时空数据的长短期变化特征,易使关键信息丢失。为此,提出一种基于LSTM-MSTCN-XGBoost混合模型的时空数据特征挖掘方法。用OWL对时空数据进行形式化建模,由LSTM与MSTCN模型分别挖掘长短期特征,输入XGBoost模型融合并输出特征模式识别结果。实验结果表明,所提方法提取的时空数据特征全局时空Moran′s I指数超过0.9,在交通时空数据挖掘中对拥堵特征的刻画也更贴合实际,可为时空数据挖掘及智能决策提供有效途径。
Due to the spatiotemporal correlation and dynamic evolution of spatiotemporal data,feature mining is difficult.The single dimensional analysis methods are difficult to comprehensively capture its long-term and short-term variation characteristics of spatiotemporal changes,which can easily lead to the loss of key information.Therefore,a spatiotemporal data feature mining method based on the LSTM-MSTCN-XGBoost hybrid model is proposed.The OWL is used to conduct the formal modeling of spatiotemporal data,LSTM and MSTCN models are used to mine long-term and short-term features respectively,and the XGBoost model is input to fuse and output feature pattern recognition results.The experimental results show that the spatiotemporal data features extracted by the proposed method have a global spatiotemporal Moran′s I index exceeding 0.9.In traffic spatiotemporal data mining,the characterization of congestion features is also more realistic,providing an effective approach for spatiotemporal data mining and intelligent decision-making.
作者
李阳政
易吉良
LI Yangzheng;YI Jiliang(Hunan University of Technology,Zhuzhou 412000,China;Guilin University of Aerospace Technology,Guilin 541004,China)
出处
《现代电子技术》
北大核心
2025年第16期157-160,共4页
Modern Electronics Technique
基金
广西自然科学基金项目(2024JJA160324)。
