Spatial Data Mining of Colocation Patterns for Decision Support in Agriculture
Han-Wen Hsiao, Meng-Shu Tsai, Shao-Chiang Wang · 2006
Computer technologies have been introduced into the area of agriculture recently. Precision agriculture, as an example, is a popular concept of using GIS, GPS and other new technologies in helping farmers optimize agricultural production. Colocation pattern mining is a technique for discovering relationships between different thematic features in a spatial domain. For example, an observation that large cities are often close to riversides is obtained with a reliable statistic. Such desired capability is of importance in agricultural applications, like insect pest management. In this paper, a two-phase hierarchical clustering method is proposed to assist people in making decisions based on spatial colocation patterns implicitly existing inside the geographical data sets. It is designed to be a generic system for any data sets in point format. In the first phase, the point features being close together are grouped into a number of clusters. An LC matrix is generated to describe the relationship between the clusters and the layers of feature points. The LC matrix is then analyzed by the second hierarchical clustering to generate a dendrogram. The support and confidence of each single cluster in the dendrogram are calculated to show the concurrent occurrence of features, regardless of their geographical locations. Key words: spatial colocation pattern, hierarchical clustering, decision support. 1.