An Probabilistic Method Of Mining Association Rule From Spatial And Temporal K-Anonymity Datasets

With the rapid development of the Internet, mobile positioning and other technologies, location-based services(LBS) has been widely used, resulting in a large amount of mobile data with temporal and spatial attributes.In the back of a large number of data related to a variety of privacy, with the improvement of people’s awareness of privacy protection and the k- Anonymity Privacy protection model is becoming more and more perfect, the space-time K- anonymous technology has become a mainstream direction of LBS privacy protection research.The Kanonymous data has the characteristics of probability and distribution. The traditional data mining method and the uncertain data mining method can not be applied to the K- anonymous data effectively.Mining Spatial-temporal K- anonymity in order to achieve its usability becomes an urgent problem to be solved.Association rule is an important type of knowledge in anonymous set data. This paper mainly studies the method of mining association rules from LBS snapshot query:(1)On the basis of the temporal and spatial k-anonymity practical significance,a transaction will be single snapshot query anonymity set is mapped to the association rule mining and based on the transaction of the components that anonymity set data(user, cell and time)elements with uncertain characteristics, design the anonymity set data item set support degrees of probability calculation method.(2)Further considering the characteristics of uneven distribution of the temporal and spatial K- anonymous data, set different support thresholds for different combinations of itemsets.In the process of generating frequent itemsets, user, cell, and time are set with different support thresholds.<user, cell>, <user, time>, <cell, time> are set different support thresholds in the process of generating frequent two items.In the process of generating frequent three items <user,cell,time > by the frequent two items are also set different support thresholds.(3)By using the Apriori characteristics of association rules, all the data association rules are generated by the method of candidate generation and test.Finally, the experimental analysis is presented in this paper the number of anonymous set of association rules data mining algorithm with set support and confidence threshold variation, and with the traditional method in the number of rule mining and based on rules for the prediction of performance(including the recall and precision) were compared.The results show that the proposed method is more useful and can reflect the association rules of the original anonymous set data, and has good prediction performance.