A novel temporal clustering technique and quality evaluation measure for group record linkage

Research in the social sciences is increasingly based on large and complex data collections, where individual data sets from different domains need to be linked to allow advanced data analysis. A popular type of data used in such a context are historical registries containing birth, death, and marriage certificates. Once such data sets are linked, pedigrees for full populations can be constructed. These will facilitate novel studies to, for example, investigate how education, health, mobility, and employment have influenced the lives of people over several generations. In this talk I will present our recently developed novel temporal clustering approach which is aimed at linking records for a group of individuals, such as all births by the same mother, and where temporal constraints need to be enforced, such as intervals between births. We then present a novel cluster quality evaluation measure that categorizes each individual record according to the quality of the cluster the record has been linked into. Experiments on a real Scottish data set show the superiority of our novel temporal clustering approach over a previous approach for group record linkage, while also highlighting the need for novel quality evaluation measures for group record linkage. This work was conducted with Ms Charini Nanayakkara and Dr Thilina Ranbaduge. For details about the temporal techniques see: link.springer.com/chapter/10.1007/978-3-030-16145-3_41Fundación BBVAN