Research On The Algorithms Of Association Rules And Hyperclique Mining

Data mining technology has been a hot task in the areas of database and artificial intelligence in recent years, and it has attracted extensive attention in science and industry field. The main purpose of data mining is to extract implicit, previously unknown, and potentially useful information out of large amounts of collected data. Association rules originated from the analysis on the supermarket shopping cart, and are mainly used to search for relationship among items in a data set. Association rules is one of the issues that study at headmost in data mining, and also a hot research direction. Association rules can be extensively applied in many fields, it is not only can check the patterns of knowledge which are long-established in the industry, but also can find some new rules which are hidden before. One of the important research aspects of data mining is how to find effective way to discover, apprehend and apply association rules.In the background of data mining and association rules mining, this paper develops research on the method of association rules mining. First, we analyzed and discussed the emergence and development of the data mining technology, the basic process of data mining, the main task of data mining; then introduce the basic concept of association rules, the algorithm research of association rules mining, the expansion and application of association rules mining, the horizontal and vertical distribution of data set; analysis of the classical methods of association rules mining algorithm Apriori algorithm and Relim algorithm which is easily implemented.We focus on the h-confidence measures in association rules mining and discuss the concept of the h-confidence, hyperclique patterns and cross-support. Then present a way of transaction splitting in hyperclique mining to preprocess transaction database, discuss how to make transaction splitting and its. correctness prove. In fact, when we mine association rules based on interesting measures, we could use transaction splitting to preprocess the transaction database if the measure has the cross-support property.In this paper, we present the hyperclique and maximal hyperclique mining algorithm which are improved from algorithm Relim. And our experiments show that transaction splitting is an effective way to improve the hyperclique mining algorithm. Our maximal hyperclique mining algorithm has a good performance in sparse data set told by experiments.After a large number of experiments on Apriori, Relim and FP-Growth by different order of items in a transaction, and study the result of experiments, we present a balance rules in association rules mining, the more close to average of the output of frequent itemsets which is divided by its first item, the more effective of the algorithm. This rule helps us to improve existing algorithm and find new algorithm.