Discovering both positive and negative co-location rules from spatial data sets

With the explosive growth and extensive applications of spatial data sets, it is becoming more and more important to solve the problem how to discover knowledge automatically from spatial data sets. Co-location patterns discovery is an important branch in spatial data mining. Traditional algorithms for co-location patterns mining can only find positive co-location patterns. However, negative co-location patterns, which are strong negative associated but whose participation index are less than a minimum prevalence threshold, sometimes would include great valuable information. In this paper, the concept of the negative co-location patterns is defined. Based on the analysis of the relationship between negative and positive participation index, methods for negative participation index calculation and negative patterns pruning strategies are given. The methods make it possible to discover both positive and negative co-locations efficiently. The applications of the proposed algorithm are studied using the plant data sets of the "Three Parallel Rivers of Yunnan Protected Areas". Finally, an extensive experimental analysis is done to show the effectiveness and efficiency of the algorithms.