关键词: 空间数据挖掘, co-location 模式挖掘, 加权参与率, 星型邻居模型, 模式并置 值

Abstract: A co-location pattern is a subset of spatial feature set whose feature instances frequently appear in proximity in geographic space. Pattern mining is carried out based on the feature participation rate， which is defined as the radio of the number of unique instances in the pattern table to the total number of feature instances. A weighted co-location pattern mining method based on specified features is proposed to address the situation where the degree of concatenation of pattern feature instances based on specified features satisfies the pattern guidance requirements， but some features have too many total instances leading to a feature participation rate below the threshold， and the pattern is defined as a non-frequent pattern. The proposed method defines the weight of features and the calculation rules of weighted participation rate， which can effectively mine weighted co-location patterns based on specified features， and the weighted participation degree decreases monotonically with the increase of the pattern order. The experimental results indicate the effectiveness of the algorithm in terms of mining results and algorithm runtime. 

Key words: spatial data mining； co-location pattern mining； weighted participation rate； star-neighbor model； mode collocation value