Research On Algorithm For Association Rule Mining And Its Application On Railway Tunnel Safety Management

The data mining technique was born accompanied by the explosive growth for the amount of information. How to meet the needs of people who are eager for knowledge and information in order that they can acquire knowledge quickly and apply it to the decision making field, has become one of the hot topics in current information processing.There are many methods for data mining, of which the association rule mining is a data mining method that is in the lead and with the widest usage. The concept of association rule was first presented in 1993 by Dr. Rakesh Agrawal who held a postion in IBM and others in order to be used for depicting the relationships among the transaction items in transaction database, i.e. the frequent relationships. The research on it has lasted more than 10 years till now and has made great achievements. But there are still many problems in urgent need of being solved. This article introduces these in details, and goes into the association rule mining theory, especially the association rule mining algorithm, with certain achievements made.The article at first introduces the meaning, process and some methods for data mining, and then introduces in brief the present research situation at home and abroad. The association rule method is a data mining method that is proved very useful in practice (in Wal-Mart).The main task of the association rule is to obtain frequent itemsets, the classical algorithms for which are Apriori algorithm and FP-growth algorithm. From the point of view of the data structure for algorithm, these two algorithms are based on arrays and trees respectively. The article also introduces and analyzes several improved algorithms such as AprioriTid and AprioriHash, and an improved algorithm of FP-growth algorithm - PFP-growth as well. After summarizing the characteristics of these algorithms, the author presents AprioriN algorithm based on arrays. This algorithm converts the operation on database into the operation on memory via coding, enhancing the efficiency of the algorithm significantly. But it has higher demands of the processor. At the same time, the author also presents a high performance association rule mining algorithm based on FP-tree - FP-growth algorithm, which quickens the traverse speed of itemsets through adding an extra data structure. During the second-time scan of the database, a matrix Aφthat can be used for saving...