A novel multi-view similarity for clustering spatio-temporal data

Abstract

With the enhanced usage of sensors and GPS devices, obtaining spatial and spatio-temporal data has become easy and analyses of these data in real-time applications are increasing day to day. Clustering is a data mining technique used for analyzing and obtaining unknown/hidden knowledge from the data/objects. Distance-based methods are helpful for analyzing and grouping the objects. In general, based on the type of data, Euclidean or Cosine distance-based techniques are used for grouping the data. Traditional techniques are point-based techniques and are based on single-view point, which may not produce efficient information and cannot be utilized for analyzing spatio-temporal objects. Hence, this paper presents a novel multi-view similarity technique for clustering spatio-temporal objects. Authors demonstrated the effectiveness of the proposed technique by adopting DBSCAN and implementing JDK1.2 on benchmarked datasets with respect to FMI indicator.