Research On Association Rules Algorithm For Big Data

Association rule mining is an important research branch of data mining, which dig up the valuable information and knowledge. With the continuous development of database applications, increase the speed of data acquisition and storage, the traditional association rule mining algorithms are not able to mine big data association rules, so it is very necessary to improve performance of mining algorithms.Firstly, this paper analyzes the research dynamic, current status and development trends of the algorithm for large data association rules mining.Then, the classical Apriori algorithm for mining association rules is analyzed in detail.According to this algorithm's defects of scanning database for more times and produce a large number of candidate item-sets, Apriori-Hybrid and Apriori-Tid which are based on Apriori algorithm are introduced. The improved algorithms only need to scan the database once and are able to remove the non-frequent item-sets in the whole process timely. The improved algorithms have a large promotion in the efficiency of time and space, and are more suitable for big data association rules, mining. After introducing, experiments are conducted, and results show that AprioriHybrid method has a better performance.Most studies using the frequent degree and support (e.g.: frequent item-sets mining), or utility and profit (such as high efficiency mining) as measure methods. Using the Apriori algorithm and its modified association rule algorithm, the frequent itemsets are identified. We have a deep research for the effectiveness and functional in the mining process. It is concluded that Association rule mining algorithms are better than other algorithms in processing big data and keeping the algorithm accuracy. This article will use these two kinds of measures together, the FHIMA (frequent and high utility item-sets mining,) algorithm has carried on the comprehensive analysis and it is applied in PDMiner Platform and experiment results are analyzed. Results show that using a tight upper bound can make FHIMA algorithm more efficient and more meaningful.In addition, some application research for big data association mining is conducted.Compared with other methods, the improved method has better performance than other mining algorithm.