Data mining-based fragmentation of XML data warehouses

Abstract

With the multiplication of XML data sources, many XML data warehouse models have been proposed to handle data heterogeneity and complexity in a way relational data warehouses fail to achieve. However, XML-native database systems currently suffer from limited performances, both in terms of manageable data volume and response time. Fragmentation helps address both these issues. Derived horizontal fragmentation is typically used in relational data warehouses and can definitely be adapted to the XML context. However, the number of fragments produced by classical algorithms is difficult to control. In this paper, we propose the use of a k-means-based fragmentation approach that allows to master the number of fragments through its k parameter. We experimentally compare its efficiency to classical derived horizontal fragmentation algorithms adapted to XML data warehouses and show its superiority.

Dimension table driven approach to referential partition relational data warehouses

“…These algorithms control of generated fragments of a fact table. [13] proposed an approach for fragmenting XML data warehouses characterised by three main steps: (i) extraction of selection predicates from the query workload, (ii) predicate clustering and (iii) fragment construction with respect to predicate clusters. To form clusters of predicates, the authors proposed to use the k-means technique.…”

“…Based on the coding in Table 1 The fact table is then decomposed into 4 fragments using the native referential partitioning as follows: CREATE For the automatic partitioning, the DBA can use one of the existing algorithms [1,13]. The main advantages of our procedure are: (i) a reduction of the complexity of the horizontal partitioning problem and (ii) an improvement of star join query performance.…”

“…By analyzing the most important works done on referential partitioning in the data warehousing environment, we draw the two following points: (a) they control the number of generated fragments of the fact tables in order to facilitate the manageability of the data warehouse [2,13,22] and (b) they start from a bag containing all selection predicates defined in a set of frequently asked queries. Dimension tables referenced by these predicates have the same chance to be used to referential partition the warehouse which is not the case in all situations.…”

Partitioning methods for multi-version XML data warehouses

“…The authors of [23] propose a data mining-based partitioning method, where the XML query predicates are clustered, and used to produce primary horizontal dimension fragments. One fragment will correspond to one cluster of predicates.…”

A Methodology and Tool for Rapid Prototyping of Data Warehouses Using Data Mining: Application to Birds Biodiversity

“…OLAP design by DM). For example, [4,18] uses DM clustering algorithms to define hierarchies, and [12] to define physical models. In the context of conceptual modeling a lot of effort has been done for the DW design.…”