Concept Hierarchy For The Background Knowledge Of The Association Rule Mining Algorithm Analysis

Because the rapid development of the database technology and method diversification that people get data, the data we can access are increasing rapidly. But there are few tools through which we can analyze and understand these data. Things that database system can do are accessing these data and some simple operations. There are a lot of information which are useful for decision-making behind the large amount of data, but it is not an easy thing to take these useful information from the great lot of data. Data mining is a process through which people can take the useful information and knowledge from vast, incomplete and noisy data.Association analyses is the forthgoer of the data mining, and which has few content intercrossed with other subject. Apriori algorithm is the base of the association analysis, the central problem of the quantitative association rule mine algorithm is to convert the continuous value attribute association analysis into Boolean value attribute association analysis. Algorithms of multi-dimension association rule and association rule based on constraint are methods for solving practical problems. The methods of data mining include statistics analysis, classification, estimate, prediction, relevant analysis, association rule, clustering and some new methods such as decision tree, neural network, data visualization.In this paper, the origin, the function and the classification of the data mining are introduced in brief. The algorithm of Apriori is analyzed in detail, which is used to find the frequent item-sets in the data-mine process of association rules; and the shortage of this algorithm is also showed. On this base, a new association rule finding method is given, which takes the concept hierarchy as the background knowledge and includes same hierarchy algorithm, mixed hierarchy algorithm and crossed hierarchy rules mining method. The multi-dimension association rule mining method becomes from one attribute expanding to multi-attribute at the same time. So the shortage of the Apriori algorithm was overcame, and the functionality of the algorithm and scope which can be used are prominent enhanced. Finally, some expectations for future work are presented.