传统的空间关联规则挖掘,一般是使用属性关联规则的挖掘算法,对空间数据进行泛化处理,不考虑空间数据的空间自相关性,也没有考虑空间自相关与空间关联规则的关系。本文运用改进的Apriori算法对某一数据进行空间关联规则挖掘,并对同一数据进行空间自相关分析,比较两种方法反映的属性的相关性,探讨了数据的空间自相关性对空间关联规则挖掘的影响。论文采用2000年英国的HAYFEVE患病数据集和当时的气温、降雨数据作为实验数据。采用两种方法处理相同的数据集,即Apriori方法和空间自相关方法,发现二者的结果中所得的一项关联规则和二项关联规则一致,证明了通过研究数据的空间自相关性也能获得准确的关联规则,且数据的空间自相关性对关联规则的挖掘存在作用和影响。如何定量度量一元空间自相关对空间关联规则的影响,以及利用二元空间自相关结果作为空间关联规则候挖掘的候选频繁项集,进而提高挖掘效率是本文的进一步工作。

Spatial autocorrelation is a very general statistical property of spatial variables, it indicates correlation of a variable with itself through space. Spatial association rule mining, discovery of interesting, meaningful rules in spatial databases, ignores autocorrelation of spatial data, or just generalizes the spatial data into attribute data currently. In most of the ways on spatial association rules mining, they transferred the spatial relations into non-spatial relations by virtue of spatial analysis. This means the separation of spatial autocorrelation from spatial association rule mining. In order to study the relations between spatial autocorrelation and spatial association rule mining, in this paper, the spatial association rules were mined by developed Apriori algorithm. Then, spatial autocorrelation analysis was implemented in the same spatial data set. A basic assumption of many spatial association rules mining is lacking for a priori information about spatial attributes. The two dimensional spatial autocorrelation results were used as priori knowledge in spatial association rules mining in this paper. The experimental data is about the amount of the hay fever (disease caused by pollen allergic rhinitis) patients and its factors, including temperature, precipitation and vegetation types of each county in the United Kingdom in 2000. The obtained frequent itemsets and the spatial association rules prove that factors have stronger correlation with hay fever (correlation coefficient is lager) appear with hay fever simultaneously more frequently in the spatial database, which confirms the existence of the effects that spatial autocorrelation has on spatial association rule mining. The analysis results not only point out the relation between spatial autocorrelation and spatial association rule mining, but also provide priori knowledge in the process of spatial association rule mining, making the mining process more targeted. Besides, without calculating the Cartesian in developed Apriori algorithm, spatial autocorrelation analysis can get the correlation coefficients efficiently, making the mining process more effectively. Further work would focus on how to evaluate the effects of the spatial autocorrelation on spatial association rules mining, how to find out the candidate frequent spatial itemsets from the results of spatial autocorrelation analysis in practical application.