Application Research Of Association Rules In Cross Selling Of Banks

Association rule is one of the most important research directions in data mining. With the continuous development of the internet and data storage technology, association rule is getting more attention and wide used in various fields. Apriori is the classic algorithm in the mining of associate rule. It was proposed by R.Agrawal in 1994 and achieved by an iterative method based on layer search, which contains two steps of connection and pruning. The algorithm is simple and easy to implement, which can get different associate rules by different support degree and confidence degree. However there are some disadvantages of Apriori algorithm. A large number of candidate sets will be generated in the connection step, thus the database need to be scanned multiple times to calculate the support level, which will affect the efficiency and adaptability of the algorithm when the data itself is big or the amount of the association rules is master.In this paper, we have studied the association rules and Apriori algorithm based on the background of the cross selling in banks. And finally we get an improved combination algorithm. The algorithm mainly consists of two parts, which are data preprocessing and generation of association rule. The first of data preprocessing should do is to process the data which is incomplete, noisy and inconsistent into another form that meets the standards of data mining. The basic methods of data preprocessing are introduced in this paper, due to the data we used here has been processed, what we should do is to make the data become discrete and digital according to the characteristics of the property of the data. Data normalization is only part of the data preprocessing, and in the big data set, not all information is useful. Then the next we will introduce an iterative algorithm based on the approximate mass of collections to reduce the attribute set. In every iteration, we should ensure the approximate mass is not smaller than before, until we get the smallest reduction sets. During the generation of association rules, we used a method of vertical format data distribution, which were added an idea of greedy in this paper. In this way, the number of intersection in the process of getting frequent item sets is reduced, accordingly, the storage is reduced.The main advantage of the algorithm in this paper is to obtain the association rules with the smaller attribute set and less storage space, which doesn't affect the ability of classification decision. It's obvious in the analysis of the experimental results that there is no difference in the number of the association rules between the algorithm in this paper and the Apriori algorithm, while in the efficiency of the execution time, the algorithm is better than Apriori algorithm.