Privacy protected mining using heuristic based inherent voting spatial cluster ensembles

Abstract

Spatial data mining i.e., discovery of implicit knowledge in spatial databases, is very crucial for effective use of spatial data. Clustering is an important task, mostly used in preprocessing phase of data analysis. It is widely recognized that combining multiple models typically provides superior results compared to using a single, well-tuned model. The idea of combining object partitions without accessing the original objects’ features leads us to knowledge reuse termed as cluster ensembles. The most important advantage is that ensembles provide a platform where vertical slices of data can be fused. This approach provides an easy and effective solution for themost haunted issue of preserving privacy and dimensionality curse in datamining applications.We have designed four approaches to implement spatial cluster ensembles and have used these for merging vertical slices of attribute data. In our approach, we have brought out that by using a guided approach in combining the outputs of the various clusterers, we can reduce the intensive distance matrix computations and also generate robust clusters. We have proposed hybrid and layered cluster merging approach for fusion of spatial clusterings and used it in our three-phase clustering combination technique. The major challenge in fusion of ensembles is creation and manipulation of votingmatrix or proximity matrix of order n2, where n is the number of data points. This is very expensive both in time and space factors, with respect to spatial data sets. We have eliminated the computation of such expensive voting matrix. Compatible clusterers are identified for the partially fused clusterers, so that this acquired knowledge will be used for further fusion. The apparent advantage is that we can prune the data sets after every (m−1)/2 layers. Privacy preserving has become a very important aspect as data sharing between organizations is also difficult. We have tried to provide a solution for this problem.We have obtained clusters.