Research On The Association Rule Algorithm And Parallelization

The Discovering Association Rules is an important research topic in data mining domain, which may find certain specific rules in the magnanimous data. The association rules can assist the policy-maker to carry on the decision, which plays an important role in the analysis of stategy, the analysis of web log, the personalization information recommendation, the market investigation and so on.In the algorithms research of Discovering Association Rules, it mainly includes two parts, one part is data mining of the frequent item sets, and other part is the generation of the association rule. The data mining of frequent item sets decides the generation of the association rule. Through the data mining of frequent item sets, we may find the correlational dependence between each item from the massive transactions. In the frequent item data-mining algorithm developing process, the Apriori algorithm and the FP-Growth algorithm are two quite typical algorithms. This paper has referred to the related compression tree's concept, has proposed the TreeMatrix algorithm compresses transactions to a compression tree and uses the matrix mode to store compression transaction sequence, then utilises stack to carry on the valid mining. On the foundation of algorithm, we have taken massive optimized measures. These measures include tree's layer upon layer mining, the frequent item of set size limit and so on, which play the quite essential role in the serial algorithm efficiency. The experiment shows the TreeMatrix algorithm and its improvement can efficiently enhance the efficiency of mining of the frequent item of sets. This paper has also conducted the parallel research of the TreeMatrix algorithm, realized the parallel process of the part of mining. The experiment indicates it has achieved the preliminary parallel purpose.