Impact of variations in anonymous record linkage on weight distribution and classification.

Anonymous or privacy preserving record linkage is the term for systems allowing the linkage of data from different sources while maintaining an individual's anonymity. This work displays the impact of variations in the process of generating weights in a probabilistic record linkage system on different datasets, the resulting set of weights of candidate pairs and consequently on the final classification process. Furthermore, the results give insight into general problems of current unsupervised classification methods.