A framework for co-location patterns mining in big spatial data

Abstract

Modern big data platforms such as Apache Hadoop and Apache Spark are able to process and analyse huge data sets, but still lack comprehensive support for spatial data analysis. Nevertheless, spatial data mining requires an efficient distributed processing of big spatial data. Spatial data mining is a subclass of data mining, which mainly focuses on obtaining explicit knowledge, spatial relations and interesting patterns from spatial data. Co-location pattern mining is one of the spatial data mining challenges. Spatial co-location pattern could be defined as a set of spatial objects or relationships which are frequently observed together in a spatial proximity. This work is mainly focused on development of a framework for co-location patterns mining in big spatio-temporal data. We also make evaluation of applied algorithms from the point of their efficiency and scalability.