This work aims to maximise the utility of published data for the partition-based anonymisation of transactional data. We make an observation that, by optimising the clustering i.e. horizontal partitioning, the utility of published data can significantly be improved without affecting the privacy guarantees. We present a new clustering method with a specially designed distance function that considers the effect of sensitive terms in the privacy goal as part of the clustering process. In this way, when the clustering minimises the total intra-cluster distances of the partition, the utility loss is also minimised. We present two algorithms DocClust and DetK for clustering transactions and determining the best number of clusters respectively.
|Title of host publication||Pacific-Asia Conference on Knowledge Discovery and Data Mining|
|Editors||Jinho Kim, Kyuseok Shim, Longbing Cao, Jae-Gil Lee, Xuemin Lin, Yang-Sae Moon|
|Place of Publication||Cham|
|Publisher||Springer International Publishing AG|
|Number of pages||14|
|Publication status||Published - 2017|
|Event||The Pacific-Asia Conference on Knowledge Discovery and Data Mining 2017: PAKDD 2017 - Seogwipo KAL Hotel , Jeju, Korea, Republic of|
Duration: 23 May 2017 → 26 May 2017
|Conference||The Pacific-Asia Conference on Knowledge Discovery and Data Mining 2017|
|Country||Korea, Republic of|
|Period||23/05/17 → 26/05/17|
Bewong, M., Liu, J., Liu, L., & Li, J. (2017). Utility aware clustering for publishing transactional data. In J. Kim, K. Shim, L. Cao, J-G. Lee, X. Lin, & Y-S. Moon (Eds.), Pacific-Asia Conference on Knowledge Discovery and Data Mining (pp. 481-494). Springer International Publishing AG.