Identifying suitable attributes for Record Linkage using Association Analysis

Record Linkage is the process of consolidating data from several sources and capturing the records that are associated with the same entities, or individuals, where a unique identifier is not available. Record Linkage has applications in several areas such as master data management, law enforcement, health care, social networking, and historical research. Identifying the attributes from the data that are best suitable for performing record linkage in terms of accuracy and linking time is crucial. A straightforward approach of using all the attributes will result in large run times and may not yield a high accuracy. It is thus important to find a subset of the attributes that would give a high accuracy, while also minimizing the linking time. In this paper, we introduce a novel idea of addressing this problem using association analysis. We describe a pipeline to process the data and an algorithm based on association rules mining to find good features using some known rule quality measures. Our experimental results reveal that our approach yields better accuracy and run times than previous algorithms.