局部空间同位模式在作物种植适宜性调查中的应用

Application of regional co-location patterns on planting suitability surveying

空间同位模式的局部挖掘算法存在挖掘结果无统计显著性意义、缺少领域知识介入导致挖掘结果针对性不强等问题.通过引入领域专家知识选择候选模式,构建候选模式属性的兴趣度函数作为局部空间统计量,对空间候选模式的兴趣度值聚集区域进行挖掘,进而合并聚集区域的邻域确定模式的空间区域.经过分析及验证,利用该方法对长汀县烟草种植适宜性结果数据集进行处理,发现了主要养分高但不适合种植烟草的异常模式及区域.据此,专家有针对性地提出了加大灌溉基础设施建设、施用石灰和硼肥等改善烟草种植条件的建议.

To solve problems of current mining algorithm for regional co-location pattern that lacks in statistical significance and pertinence,a new approach based on experts＇ interestingness was proposed. Firstly,co-location pattern consisting of sets of continuous variables was screened and mined by the function of experts＇ interestingness. Hotspot and target spacial region of interests was extracted and integrated from co-location pattern. Furthermore,the approach was validated by applying to tobacco suitability evaluation data sets of Changting County in Fujian Province. Abnormal region in Changting that had fertile soil but was unsuitable for tobacco cultivation was identified. Suggestions including emphasizing irrigation infrastructure and application of lime and boron fertiliser were put forward.