Association Rule Mining Algorithms, Data Mining Technology

Data mining accompanied by the rapid growth of information, it was born in the process of data disjunctive, identification and then finding potential, useful, hitherto unknown, ultimately comprehensible knowledge (rules or model) in it. In order to help users understand the existing information, the data mining technology should base on the existing data to identify the data model, and make a forecast to future situation on the basis of the existing information.Mining of association rules is an important research in data mining, mainly used for data concentration items which related with each other. Apriori algorithm and FP-Tree algorithm is the classical algorithms in the association rule mining, they are based on the transaction database data will not be changed and each data items are equal importance. But the data in the actual application of the database is changing constantly, and people show different degree of attentions towards the data items. If we still use the traditional mining algorithm for association rules mining, the mining efficiency will be very low, and the results are not accurate. In order to solve these problems, this thesis gives an in-depth study, the main work and innovation of this work are as follows:(1) The Boolean type algorithm for mining association rules algorithm—Apriori algorithm is studied, and the basic ideas of the algorithm and its mining steps are summarized. All kinds of improved measures to this algorithm are discussed, according to the detailed analysis of the shortcomings of the algorithm. Meanwhile, we make a detailed algorithm analysis and research with the influential improved algorithm—FP-Growth algorithm.(2) A improved algorithm for mining incremental updating algorithm—AFUP algorithm is proposed. According to the current incremental updating algorithm is not sensitive to the new projects, the concept of sensitivity to measure the sensitive degree of new projects is introduced in this algorithm. The problem of the traditional incremental updating algorithm can't discover the concentrated potential association relationship of new projects can be solved by AFUP algorithm, thus the efficiency of the algorithm improved.(3) An new kind of weighted association algorithm for mining association rules algorithm—FPWAL algorithm is proposed. Different with the mining algorithm in the basis of the Apriori algorithm, FPWAL algorithm based on the FP-Tree algorithm, reducing the number of scanning databases, greatly improving the mining efficiency. In addition, the concepts of the vertical and horizontal weights are introduced in this thesis, it can make mining results more reasonable.In this thesis, we have both two algorithms proved, and the results show the effectiveness the rationality and high efficiency of the proposed algorithm. At the same time, the proposed algorithm can also made the results of mining more realistic demand.