Discovering the geographical connection between events provides managers and planners with valuable knowledge that can be used to make better decisions. There are countless methods for investigating the geographical connection of phenomena, among which we can mention co-location. Finding co-location patterns as a type of relationship between places, infers the existence of one set of spatial features based on the existence of another set of features. Co-location patterns in spatial data sets are a set of dissimilar spatial objects whose samples are often located in a geographical area close to each other. Extracting these patterns can lead to the discovery of interesting relationships that provide important insights for various applications such as public health, epidemiology, environment, criminology, transportation, and various location-based services. Some co-location patterns appear globally in the entire study area, which are called global co-location patterns, while others can only be seen in parts of the study area, which are called local co-location patterns. become The discovery of multilevel co-location patterns provides new insight into the interaction between different spatial phenomena and has wide application in geographic domains. In recent years, due to the importance of multi-level co-location patterns, many researches have been conducted, but still discovering multi-level co-location patterns is a challenging task. The limitation of most of the existing algorithms is that they work with point spatial features and a single neighborhood distance threshold, while in reality there are datasets that have lines and polygons in addition to point samples, and not paying attention to this issue may result in to affect the obtained. Also, spatial concepts can be categorized under a higher level concept, and moving along these hierarchies leads to the discovery of patterns that may not be noticed normally. Because it is possible that there are co-location patterns that were ignored at a higher granularity level and are considered important at a lower granularity level. In order to mining local patterns at different levels of granularity, in this research, a part has been added to the base algorithm for co-location pattern mining, which is able to enrich the non-point data set of the studied area, such as the neighborhood data set of that area and perform pattern mining based on granularity, at two levels of granularity, medium and fine.. The results of running the model on spatial datasets show that by using this model and by solving the stated problem, patterns with high preva‎lence and more meaningful results are obtained

ناصر قديري مدرس

محمدرضا احمدزاده، عليرضا بصيري