An Algorithm about Association Rule Mining Based on Spatial Autocorrelation

Most spatial data in GIS are not independent, they have high autocorrelation. For example, temperatures of nearby locations are often related. Most of the spatial association rule mining algorithms derived from the attribute association rule mining algorithms which assume that spatial data is independent. In these situations, the rules or knowledge derived from spatial mining will be wrong. It is, therefore, important that mining spatial association rules take into consideration spatial autocorrelation. At present, spatial statistics are the most common method to research spatial autocorrelation. In spatial statistics, classic statistics are extended by taking into account spatial autocorrelation. Spatial Autoregressive Model, SAR, is one of the methods; the adjacency matrix is used to describe the interaction of neighbouring fields which can simulate the effect of dependence between variables. The disadvantage of spatial statistics is that the calculation consuming is high so it cannot be widely applied in spatial data mining. This paper puts forward a new method of mining spatial association rules based on taking account of the spatial autocorrelation with an cell structure theory. It defines spatial data with an algebra data structure then the autocorrelation of the spatial data can be calculated in algebra. According to J. Corbett’s cell structure theory (1985), spatial graph is a subset of point, line, face, and body. The algebra structure of point, line, face and body can be used to express spatial data. In spatial data mining, we mine rules in the spatial database. In this paper the first step is to design a structure about point, line, face and body to express the spatial data and then store it in the spatial database. The second step is to build the measurement model of spatial autocorrelation based on the algebra structure of spatial data. The third step to mine the association rules based on the spatial autocorrelation model. Taking account of spatial autocorrelation is a significance research field for mining spatial association rules. We can get the spatial frequency items from the autocorrelation of the spatial data. This replaces the repeated scanning of the database by the measure of the spatial autocorrelation.