{"title":"基于自适应信息聚合的空间托普利兹反协方差的多变量时空聚类法","authors":"Ling Wang , Gang Wang","doi":"10.1016/j.neucom.2025.130172","DOIUrl":null,"url":null,"abstract":"<div><div>Spatiotemporal clustering is an important technique for discovering potential and useful patterns in many areas including public safety, ecology, epidemiology, earth sciences, etc. However, many spatiotemporal clustering methods fail to effectively capture the complex correlations between sequences in spatiotemporal data, which leads to clustering results that are spatially scattered. Hence, we propose a multivariate spatiotemporal clustering method based on spatial Toplitz Inverse Covariance with Adaptive Information Aggregation (TIC-AIA). Specifically, in the temporal dimension, we extend Gaussian Markov Random Field to extract temporal features from multivariate spatiotemporal data. In the spatial dimension, a new adaptive K-nearest neighbor algorithm is proposed to construct subregions of different sizes for each location. Then, an adaptive information aggregation mechanism is employed to integrate temporal features and spatial correlations to obtain the spatiotemporal information of subregions, which are applied to the spatial Toeplitz inverse covariance structure for clustering multivariate spatiotemporal data. Additionally, a new variable importance metric is developed to identify unimportant variables to optimize the clustering accuracy. Experimental results show that TIC-AIA outperforms existing techniques in clustering performance across various synthetic and real-world multivariate spatiotemporal datasets.</div></div>","PeriodicalId":19268,"journal":{"name":"Neurocomputing","volume":"638 ","pages":"Article 130172"},"PeriodicalIF":6.5000,"publicationDate":"2025-04-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Multivariate spatiotemporal clustering based on spatial Toplitz inverse covariance with adaptive information aggregation\",\"authors\":\"Ling Wang , Gang Wang\",\"doi\":\"10.1016/j.neucom.2025.130172\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<div><div>Spatiotemporal clustering is an important technique for discovering potential and useful patterns in many areas including public safety, ecology, epidemiology, earth sciences, etc. However, many spatiotemporal clustering methods fail to effectively capture the complex correlations between sequences in spatiotemporal data, which leads to clustering results that are spatially scattered. Hence, we propose a multivariate spatiotemporal clustering method based on spatial Toplitz Inverse Covariance with Adaptive Information Aggregation (TIC-AIA). Specifically, in the temporal dimension, we extend Gaussian Markov Random Field to extract temporal features from multivariate spatiotemporal data. In the spatial dimension, a new adaptive K-nearest neighbor algorithm is proposed to construct subregions of different sizes for each location. Then, an adaptive information aggregation mechanism is employed to integrate temporal features and spatial correlations to obtain the spatiotemporal information of subregions, which are applied to the spatial Toeplitz inverse covariance structure for clustering multivariate spatiotemporal data. Additionally, a new variable importance metric is developed to identify unimportant variables to optimize the clustering accuracy. Experimental results show that TIC-AIA outperforms existing techniques in clustering performance across various synthetic and real-world multivariate spatiotemporal datasets.</div></div>\",\"PeriodicalId\":19268,\"journal\":{\"name\":\"Neurocomputing\",\"volume\":\"638 \",\"pages\":\"Article 130172\"},\"PeriodicalIF\":6.5000,\"publicationDate\":\"2025-04-08\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Neurocomputing\",\"FirstCategoryId\":\"94\",\"ListUrlMain\":\"https://www.sciencedirect.com/science/article/pii/S0925231225008446\",\"RegionNum\":2,\"RegionCategory\":\"计算机科学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q1\",\"JCRName\":\"COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Neurocomputing","FirstCategoryId":"94","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S0925231225008446","RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE","Score":null,"Total":0}
Multivariate spatiotemporal clustering based on spatial Toplitz inverse covariance with adaptive information aggregation
Spatiotemporal clustering is an important technique for discovering potential and useful patterns in many areas including public safety, ecology, epidemiology, earth sciences, etc. However, many spatiotemporal clustering methods fail to effectively capture the complex correlations between sequences in spatiotemporal data, which leads to clustering results that are spatially scattered. Hence, we propose a multivariate spatiotemporal clustering method based on spatial Toplitz Inverse Covariance with Adaptive Information Aggregation (TIC-AIA). Specifically, in the temporal dimension, we extend Gaussian Markov Random Field to extract temporal features from multivariate spatiotemporal data. In the spatial dimension, a new adaptive K-nearest neighbor algorithm is proposed to construct subregions of different sizes for each location. Then, an adaptive information aggregation mechanism is employed to integrate temporal features and spatial correlations to obtain the spatiotemporal information of subregions, which are applied to the spatial Toeplitz inverse covariance structure for clustering multivariate spatiotemporal data. Additionally, a new variable importance metric is developed to identify unimportant variables to optimize the clustering accuracy. Experimental results show that TIC-AIA outperforms existing techniques in clustering performance across various synthetic and real-world multivariate spatiotemporal datasets.
期刊介绍:
Neurocomputing publishes articles describing recent fundamental contributions in the field of neurocomputing. Neurocomputing theory, practice and applications are the essential topics being covered.