Interactive Spatial Co-location Pattern Mining Based On Machine Learning

Spatial data mining is the process of discovering interesting patterns and relationships from spatial data.Co-location pattern mining,which is a type of spatial data mining,focuses on finding patterns in which two or more spatial objects frequently co-occur in close proximity to each other.These patterns are useful in various applications,such as urban planning,environmental analysis,and healthcare.One of the challenges in co-location pattern mining is dealing with the huge number of discovered patterns.With the increasing volume of spatial datasets,the number of co-location patterns that can be discovered grows rapidly.However,not all of these patterns may be useful or interesting to users,which makes it difficult to extract valuable insights from the data.To overcome the above limitations,this article proposes a machine learning-based interactive co-location pattern mining method.The specific contents are as follows:(1)To address the problem of selecting co-location patterns as dissimilar as possible for interaction with users in interactive systems,a hierarchical filtering algorithm based on colocation patterns is proposed.By analyzing the properties of spatial co-location patterns and relevant machine learning models,a unique metric formula is designed,and hierarchical filtering is used to achieve the goal of selecting dissimilar isomorphic patterns.(2)To address the issues of existing mining algorithms not considering user preferences comprehensively enough,a machine learning-based interactive mining system is proposed.a machine learning model is introduced into the interactive system to enable the system to effectively mine patterns of interest for different users.In the interactive system,users only need to label a small subset of co-location patterns as interesting or uninteresting to complete the labeling of the machine learning model training set.After the machine learning model training is complete,it can output users’ preferences(interested or uninterested)for all input co-location patterns.After several interactions,the prediction accuracy of the system will stabilize,and the interaction will stop.With the aim of facilitating user operation,a machine learning-based interactive spatial co-location pattern mining prototype system has been developed.This system is simple to operate,has a concise interactive interface,making it easy for users to use,and the prototype system can clearly display the patterns that users are interested in.The proposed method is evaluated using real and synthetic datasets,and the results show that the interactive method achieves 80% prediction accuracy in discovering user preferences for co-location patterns.This method can help users discover more valuable and interesting patterns,thereby promoting better decision-making in spatial data mining tasks.In addition,this method can be extended to other types of spatial data mining tasks and applied in various scenarios,such as location-based services,transportation planning,and marketing analysis.