Existing parallel mining algorithms for frequent itemsets lack a mechanism that enables automatic parallelization, load balancing, data distribution, and fault tolerance on large clusters. As a solution to this problem, we design a parallel frequent itemsets mining algorithm called FiDoop using the MapReduce programming model. To achieve compressed storage and avoid building conditional pattern bases, FiDoop incorporates the frequent items ultrametric tree, rather than conventional FP trees. In FiDoop, three MapReduce jobs are implemented to complete the mining task. In the crucial third MapReduce job, the mappers independently decompose itemsets, the reducers perform combination operations by constructing small ultrametric trees, and the actual mining of these trees separately. We implement FiDoop on our in-house Hadoop cluster. We show that FiDoop on the cluster is sensitive to data distribution and dimensions, because itemsets with different lengths have different decomposition and construction costs. To improve FiDoop's performance, we develop a workload balance metric to measure load balance across the cluster's computing nodes. We develop FiDoop-HD, an extension of FiDoop, to speed up the mining performance for high-dimensional data analysis. Extensive experiments using real-world celestial spectral data demonstrate that our proposed solution is efficient and scalable.Index Terms-Frequent itemsets, frequent items ultrametric tree (FIU-tree), Hadoop cluster, load balance, MapReduce.

No abstract

No abstract

Frequent itemsets mining (FIM) as well as other mining techniques has been being challenged by large scale and rapidly expanding datasets. To address this issue, we propose a solution for incremental frequent itemsets mining using a Full Compression Frequent Pattern Tree (FCFP-Tree) and related algorithms called FCFPIM. Unlike FP-tree, the FCFP-Tree maintains complete information of all the frequent and infrequent items in the original dataset. This allows the FCFPIM algorithm not to waste any scan and computational overhead for the previously processed original dataset when new dataset are added and support changes. Therefore, much processing time is saved. Importantly, FCFPIM adopts an effective tree structure adjustment strategy when the support of some items changes due to the arrival of new data. FCFPIM is conducive to speeding up the performance of incremental FIM. Although the tree structure containing the lossless items information is space-consuming, a compression strategy is used to save space. We conducted experiments to evaluate our solution, and the experimental results show the space-consuming is worthwhile to win the gain of execution efficiency, especially when the support threshold is low.INDEX TERMS Frequent itemsets mining, incremental mining, FP-tree, FCFP-tree, association rule.

For high-dimensional and massive data sets, a relevant subspace based contextual outlier detection algorithm is proposed. Firstly, the relevant subspace, which can effectively describe the local distribution of the various data sets, is redefined by using local sparseness of attribute dimensions. Secondly, a local outlier factor calculation formula in the relevant subspace is defined with probability density of local data sets, and the formula can effectively reflect the outlier degree of data object that does not obey the distribution of the local data set in the relevant subspace. Thirdly, attribute dimensions of constituting the relevant subspace and local outlier factor are defined as the contextual information, which can improve the interpretability and comprehensibility of outlier. Fourthly, the selection of N data objects with the greatest local outlier factor value is defined as contextual outliers. In the end, experimental results validate the effectiveness of the algorithm by using UCI data sets.