Analysis of Historical Census using Graph-based Household Matching Method

Abstract

Population censuses are the most useful information for developing a country. It provides a valuable description of the state of a nation. These data can be applied for the country development planning or construction process. Linking records using population census data is the linking process of the same households from several censuses across the time. The challenges of linking census comprise un-reliable data quality, lots of common names. In ten years, a household may break into several households because of marriage or dead or movement in other households. A graph-based architecture using the unique ID to match households is presented in this paper. By using the graph-based household matching method, it can achieve single and multiple household matching and can also trace family household changes between two decades. The proposed system used the unique inhabitant ID and head ID to obtain accurate results and higher similarity. As a result, the proposed method obtains 61% of Accuracy which outperforms all other compared similarity methods.