Abstract (Translated):
This paper verified the predictive performance of probabilistic record linkage algorithms for the integration big sized real databases, evaluating the effects of the blocking key definition, as well as string metric functions and phonetic code pairing algorithms with respect to the prediction's quality and computational complexity. A bibliographical survey of the main deterministic and probabilistic record linkage methods was carried out, as well as of recent advances combining machine learning techniques and main packages and implementations available in open-source R language. The results can provide heuristics for problems of administrative records integration at national level and have potential value for the formulation and evaluation of public policies