Development of Spatio Temporal Clustering Technique

Abstract

Spatio-temporal data mining (STDM) is a process of the extraction of implicit knowledge, spatial and temporal relationships, or other patterns not explicitly stored in spatio-temporal databases. It is becoming now very important field of research as it focuses the data from not only static view point but also on time and space which is dynamic in nature. Moreover ST-Data tends to be highly self auto-correlated and hence assumption which are taken in normal/Gaussian distribution fails. Here concentration is on the Spatio-Temporal Data Mining (Clustering) Technique for Earth Observation Data. After exhaustive literature survey we gathered the knowledge of different Spatio-Temporal data types, various existing Clustering Techniques, they were compared and based on the parameters like time complexity, ability to detect arbitrary shaped cluster, ability to handle high dimensional data, detection of nested clusters etc., few of them like DBSCAN, ST-DBSCAN and OPTICS were selected. Various input parameters which are taken by these algorithms were studied and for some parameters sensitivity were also observed. Implementations of these techniques were done and results obtained were displayed using QGIS open source visualization tool and they were compared. A new algorithm is being proposed called ST-OPTICS which is modified version of existing density based technique (called OPTICS).