Family tree generation from electoral data to learn geographical distribution patterns

In this era of nuclear families, each individual would hardly identify his family members beyond a couple of generations. In comparison to earlier times the volume of people who relocate to other geographical distributions for reasons related to education, work is immensely large. This in turn leads to larger disconnectivity. An attempt to build family trees as large as possible would enable us to derive useful insights about the transformations in the society. The importance of building a family tree can be highlighted from its applications ranging from tracing history of hereditary diseases to tracing criminal behavior likelihood in a suspect. The paper proposes an optimized algorithm for generating a family tree from electoral data corresponding to a geographical constituency. The major challenge imposed by the data is a non-unique foreign key to establish relationships between individuals. The paper proposes an efficient algorithm which tries to overcome the challenges imposed by the data. The model also gives a measure of the relationship strength which is an indicator of the concreteness of the link established. The results are analyzed to arrive at distribution patterns and justifiable inferences have been derived.