Improving our knowledge of metaheuristic approaches for cell suppression problems

Staggemeier, A., Thompson, Jonathan Mark, Smith, J. and Clark, A. 2007. Improving our knowledge of metaheuristic approaches for cell suppression problems. Presented at: UNECE/Eurostat Work Session on Statistical Data Confidentiality, Manchester, UK, 17-19 December 2007. Proceedings of the Eurostat Conference on Statistical Data Confidentiality. Titchford: Eurostat, 10.2901/Eurostat.C2007.004

Full text not available from this repository.

Abstract

This work will discuss further tests carried out using a pool of operational research and artificial intelligence techniques to solve the cell suppression problem. Existing solutions to the problem available through Tau-Argus software are mathematically demanding and only enable a solution to the problem for small table sizes. The approaches investigated here are pseudo-optimum in quality but enable handling of large size tables with complex structure. The test bed for this work used artificially created data which represent real-world scenarios found at ONS and a sample of real data created from IDBR sources. These data type are summarised as magnitude data, in both hierarchical and non-hierarchical formats, including 2 sets of sensitivity (2% and 10% sensitivity) and 2 sets of sparsity measures (5% and 25% of the table’s cells contain zero values). Among the approaches discussed in this paper are: a hybrid Evolutionary Algorithm, Ant Colony Optimization and Greedy Randomising Adaptive Search Procedure (GRASP). The table safety criterion was met using the Attacker Model by Salazar and Fischetti (2001). A relaxed feasibility criterion was also used on the Ant Colony and GRASP approaches in order to try to accelerate the evaluation process. Initial results showed that all approaches are able to handle larger data sets than existing mathematical programming routines. However a trade-off analysis between time taken to solve and data size indicated that we still have to improve the total time, perhaps by using not a single cell pass for table safety evaluation but a multiple cells at a time..

Item Type:	Conference or Workshop Item (Paper)
Book Type:	Edited Book
Date Type:	Publication
Status:	Published
Schools:	Mathematics
Subjects:	Q Science > QA Mathematics
Uncontrolled Keywords:	Tau-Argus, Statistical Disclosure Control, Cell Suppression, Mathematical Programming, Evolutionary Algorithms, Ant Colony Optimization, Greedy Randomising Adaptive Search Procedure (GRASP)
Publisher:	Eurostat
Last Modified:	04 Jun 2017 04:09
URI:	https://orca.cardiff.ac.uk/id/eprint/33664

Actions (repository staff only)

Edit Item