ISAMM Institute

Association rule mining is one of the most important data mining techniques. Algorithms of this technique search a large space, considering numerous different alternatives and scanning the data repeatedly. Parallelism seems to be the natural solution in order to be able to work with industrial-sized databases. Large-scale computing systems, such as Grid computing environments, are recently regarded as promising platforms for data and computation-intensive applications like data mining. However, to improve the performance and achieve scalability by using these heterogeneous platforms, new data partitioning approaches and workload balancing features are needed. The focus of this paper is to propose a dynamic load balancing strategy for parallel association rule mining algorithms in the context of a Grid computing environment. This strategy is built upon a distributed model which necessitates small overheads in the communication costs for load updates and for both data and work transfers. It also supports the heterogeneity of the system and it is fault tolerant.

Applying classical association rule extraction framework to dense SAGE data leads to an unmanageably highly sized association rule sets-compounded with their low precisionthat often make the perusal of knowledge ineffective, their exploitation time-consuming, and frustrating for the user. To overcome such drawback, we advocate the extraction and exploitation of compact and informative generic basis of association rules. Obtained preliminary results highlight that the extracted correlations may be of help in identifying a gene functional groups and thus contribute to their annotation. From a biologic point of view, such identification may be a powerful verification technique for hampering gene mis-annotating or badly clustering in the Unigene library.

In this paper, we present the Weka4GML architecture, a new framework based on WSRF technology for developing meta-learning methods to deal with datasets distributed among Data Grid. This framework extends the Weka toolkit to support distributed execution of data mining methods, like meta-learning. The architecture and the behavior of the proposed framework are described in this paper. We also detail the different steps needed to execute a meta-learning process on a Globus environment.

One of the biggest challenges for nowadays biologist is the identification, characterization and gene-gene interactions detection for common human diseases such as cancer and diabetes. This challenge is partly due to the explosion of biological information. The multifactor dimensionality reduction (MDR) method can be used to address this problem. This method can be computationally intensive, especially when more than ten polymorphisms need to be evaluated. The Grid is a promising architecture for genomics problems providing high computing capabilities. In this paper, we describe a framework for supporting the MDR method on Grid environments. This framework helps biologists to automate the execution of multiple tests of gene-gene interactions detection.

The number of Web Usage Mining (WUM) applications is growing continuously, especially due to the business interest in e-commerce Web sites and the related Web-marketing applications. The application of WUM results goes beyond the subject of our thesis since one important part of our thesis deals with the problem of selection of webviews using technics of WUM techniques. View materialization is an important issue if we want to improve the efficiency of many applications like OLAP, Database and Web applications. In this paper, we propose a novel approach for selecting webviews to be materialized in order to optimize the response time of web queries. The selection of webviews to be materialized was mainly based on the estimation of metrics requiring hard collects of multiple statistics [9] that's why, we believe on a solution based on mining an interesting set of webviews to be materialized from realistic data : Web log files. Thus, Web log files will be parsed, analyzed and treated to give a set of webviews, based on frequent closed itemsets.

The emergence of Grid architectures brings in new challenges in terms of high latencies, dynamic resource availability, and data sharing. Data Grids connect collections of thousands of geographically distributed computational and storage resources, and promote their sharing between users. Data sharing can be accomplished through data replication. Although this later is a well-established technique, its implementation presents additional problems such as the management of replica consistency. In this paper, we propose a new model to address this specific problem in large scale systems. The model combines two existing approaches. It benefits from their advantages and offers more. It optimizes the cost of replica consistency management and resists to changes in large systems.

To avoid obtaining an unmanageable highly sized association rule sets-compounded with their low precision-that often make the perusal of knowledge ineffective, the extraction and exploitation of compact and informative generic basis of association rules is a becoming a must. Moreover, they provide a powerful verification technique for hampering gene mis-annotating or badly clustering in the Unigene library. However, extracted generic basis is still oversized and their exploitation is impractical. Thus, providing critical nuggets of extra-valued knowledge is a compellingly addressable issue. To tackle such a drawback, we propose in this paper a novel approach, called EGEA (Evolutionary Gene Extraction Approach). Such approach aims to considerably reduce the quantity of knowledge, extracted from a gene expression dataset, presented to an expert. Thus, we use a genetic algorithm to select the more predictive set of genes related to patient situations. Once, the relevant attributes (genes) have been selected, they serve as an input for a second approach stage, i.e., extracting generic association rules from this reduced set of genes. The notably decrease of the generic association rule cardinality, extracted from the selected gene set, permits to improve the quality of knowledge exploitation. Carried out experiments on a benchmark dataset pointed out that among this set, there are genes which are previously unknown prognosis-associated genes. This may serve as molecular targets for new therapeutic strategies to repress the relapse of pediatric acute myeloid leukemia (AML).

Abstract—In this paper, we propose an adaptation of the Patricia-Tree for sparse datasets to generate ,non ,redundant ,rule associations. Using this adaptation, we can generate frequent closed itemsets that are more compact,than frequent itemsets used in Apriori approach. This adaptation has been experimented,on a set of datasets benchmarks. Keywords—Datamining, Frequent itemsets, Frequent closed

We investigate the problem of optimizing distributed queries by using semijoins in order to minimize the amount of data communication between sites. The problem is reduced to that of finding an optimal semijoin sequence that locally fully reduces the relations referenced in a general query graph before processing the join operations.