Framework Of Ontology-based Spatial Co-location Rules Mining(OSCRM)

Spatial Data Mining requires more challenge than the traditional data mining. With the outbreak of the data age, magnitude of spatial data also showed an increasing exponentially. Thus, the rules generated by the general rule mining methods may not be interested to users, and, with the increasing amount of data, more rules will be useless. In this case, even if there are many ways to reduce the rules, such as item-set concise representation, redundancy reduction and multiple mining, mining data based on statistical data may still not interesting. Therefore, we need some effective algorithms to help decision makers to reduce the number of generated rules.In this paper, Ontology-based Spatial Co-location Rule Mining (OSCRM) is proposed to help decision-makers get more efficient and interesting rules. First, We use ontology to aggregate users’knowledge which can express a knowledge system, ontology allows high-level data mining, thus it greatly reduces the amount of rules and time-consuming; secondly, we extend the relationship description language in order to get the user’s expectations pattern more quickly, we use marked-up language to express structured data and its relationship, meanwhile, we define some operations on relation description language, which allow users to filter more complex rule schemas, operations and description together then form a formula. Next we use Join-based and Join-less based on ontology to mine appropriate rules which are rough ones, however, the number of rules may be very large, and finally, we need to reduce the number of rules. This paper proposes two methods to filter rules-MICF and IRF, and eventually rules after reduction presented to the user. Form the above description, the entire framework is an interactive mining process.We present the conclusion and expectation of future work in the last of this paper.