Ontology-Based Approaches For Interactively Mining Co-location Patterns

Spatial data mining is the process of automatically discovering useful information from large spatial data sets.Spatial co-location pattern mining is the process of mining interesting,previously unknown but potentially useful patterns from spatial data sets.As an important branch of spatial data mining,co-location pattern mining has become a hot topic at home and abroad,and it has been widely used in fields such as ecology,public safety,transportation,and location-based services.With the development of technologies such as GPS(Global Positioning System)and remote sensing,the scale of spatial data sets is increasing.This growing spatial data imposes more severe requirements on co-location pattern mining.On the one hand,the performance of the traditional co-location pattern mining algorithm islimited.In the modern of Big Data,it is important to find a fast and effective co-location pattern mining algorithm.On the other hand,with the increase of spatial features(spatial object types),the traditional co-location pattern mining algorithms generate numerous prevalent co-location patterns.Therefore,in recent years,many approaches have been proposed to reduce the number of prevalent co-location patterns.However,they cannot guarantee that the co-location patterns that the user is really interested in can be obtained,and the patterns of interestingness to each user differ depending on their domain knowledge.In order to effectively solve above problems,we use the powerful expressive ability of ontology to domain knowledge,and through a limited number of interactions with users,discover co-location patterns that users are really interested in.The main contributions are as follows:1.The interactive mining algorithms proposed both are post-mining processes,i.e.,using a set of co-location patterns as the inputs to discover interesting co-location patterns.In order to ensure the simplicity of the input set,we propose a new set of co-location patterns called concise co-location patterns,and present a clique-based framework to discover them.The framework is more efficient and flexible than traditional co-location pattern mining frameworks.The C-hash structure constructed stores the distribution of each co-location pattern in a compressed manner,it can be used to generate concise co-location patterns quickly and efficiently.2.An interactive co-location pattern mining algorithm based on pattern genericity consisting of ontology concepts and operators is proposed.Pattern genericity represents a certain type of co-loaction patterns.The algorithm packs pattern genericity as formulas to express interesting co-location patterns,which can quickly and accurately obtain satisfied co-location patterns.3.An interactive co-location pattern mining algorithm based on pattern similarity is proposed.The algorithm uses ontology to express the domain knowledge of users,and defines the similarity between co-location patterns based on ontology concepts.The algorithm provides a small amount of sample co-locations to the user per round.The sample co-location patterns(usually with numbers of 5-10)are handed over to the user for preference selection,and after a finite round of interaction,the interesting co-location patterns are obtained.In order to reduce the number of interactions,an efficient sample co-location pattern selection method and two methods for updating candidate co-location patterns are designed.Finally,an effective filter is given to further reduce the scale of the result set.4.An interactive business site selection system based on spatial co-location patterns and ontologies is presented in this paper.The system takes the POI(Point of Interest)data as its input.Given a business that the user wants to create,this system discovers concise co-location patterns containing this business,and then mines interesting co-location patterns from the concise co-location patterns interactively with the user,and finally,performs address analysis on each row-instance of the mined co-location patterns,and recommends the final site list to the user.The system can quickly and efficiently recommend reliable site list to the user,which saves lots of manpower and resources.