A data mining framework for environmental and geo-spatial data analysis

Mining geo-spatial data is an important task in many application domains, such as environmental science, geographic information science, and social networks. In this paper, we introduce a data mining framework, which includes pre-processing of environmental and geo-spatial data, geo-spatial data mining techniques, and visual analysis of environmental and geo-spatial data. In particular, we propose new density-based clustering algorithms to identify interesting distribution patterns from geo-spatial data, a change pattern discovery technique to detect dynamic change patterns within spatial clusters, and a post-processing technique to extract interesting patterns and useful knowledge from geo-spatial data. Our density-based clustering algorithms are based on the well-established density-based shared nearest neighbor clustering algorithm, which can find clusters of different shape, size, and densities in high-dimensional data. The post-processing analysis technique allows automatic screening of interesting spatial clusters. The change pattern discovery algorithm is able to detect and analyze dynamic patterns of changes within spatial clusters. This paper focuses on developing a framework integrating a sequence of data mining process including clustering algorithm, analysis technique and pattern changing discovery algorithm. In contrast to previous works in this area, our approaches can cluster and analyze dynamically evolved complex objects, i.e., polygons. We evaluate the effectiveness of our techniques through a challenging real case study involving ozone pollution events in the Houston–Galveston–Brazoria area. The experimental results show that our approaches can discover interesting patterns and useful information from geo-spatial air-quality data.