Research On Association Rules Mining Based On Temporal Data

Association rule which is a method in data mining focuses on determining the relation between different attributes of data sets and finding the dependent relationship of the multi-attribute that meets the threshold value of the given support and confidence degree. Such rule could be applied in designing commodity shelves, arranging inventories and assigning customers according to the purchasing mode.Traditional association rules almost do not concern the time applicability. In fact, every association rule has its time period when it is correct.Therefore, the mining association rules which impose some kind of temporal binding constraints can describe objective law better. And such rules are called as temporal association rules. Temporal data mining is a new topic in data mining field, which attracted a great deal of attention. Temporal association rule mining has become a hot spot for many scholars.At first, this thesis introduces the concept, technology and research status of data mining. Secondly, it introduces the basic theory of mining association rules and classical algorithm for mining, and analyzes the algorithms. Finally, temporal data mining and mining techniques are discussed.In the traditional data mining, in order to produce the option set, it is need to scan the affair data base. However, the affair data base used in the association rule data mining is usually extremely large and need heavily I/O load. This situation must affect the efficiency of the algorithm. Based on the facts above, this thesis brings forward two improving method for the temporal association rule.One improving method is minishing the affair data sets used in the future step by step during the scanning process. This method can make the affair data base smaller and smaller, thus the efficiency of the algorithm could be increased. The other way is changing the mining order which does the data mining in the common ways disregard timing factor then considers the timing restriction and combines optimal algorithm with high efficiency to improve the performance of the algorithm greatly. And last, the thesis analyzes the capability of the method.