The Research Of Temporal Association Rule Based On An Imporvede Representational Data Model

With the development of data raining technology, temporal information is getting attention increasingly, especially the Bitemporal database which reflects not only the historical information of the incidents but also the temporal information of Meta events in the system. Technologies of bitemporal database model and index has being mature.In most cases, not all the attributes in one record is time related. As more attention on time information, more and more researches have being carried on time related attributes. Althrough the classic bitemporal database model set the effective time and event time, the effective time and event time are bound to the record but a certain attributes of the record. In this situation, which attributes are time related is hard to finger out. When time changing causes time-related attributes changing, large record will get larger IO pressure, which is negative to efficiency. In a word, temporal data mining is hard to carry on in classic temporal database model.Due to the above reasons, an improved representational relationship model is proposed in this paper. Improved representational relationship model splite up the previous bitemporal relationship into multiple time table base on the time of the attributes. Introduced category relationship manage multiple time table in improved model. In this way the time-related data analysis works is more directly and fexible. In addition, in this paper, we describe the find, update operations in detail for the new model, including: add a new object, modify attributes, delete a object and so on.In order to verify the superiority of this new model in data analysis, the paper focuses on the application of temporal association rules in data mining. Firstly, we give out the concepts of common association rules and the classical Apriori arithmetic. Secondly, on these bases, we describe how to operate the temporal association rules arithmetic on this improved new model and we have an operation procedure in detail and give out the experiment data. The expriments show that the new model has superiority in the application areas. The inadequacies of the new model is that the model doesn't take indexing technology and TSQL2 language support , which will be researched in the future.