Multi-party record linkage method and system, electronic device and storage medium

Abstract

The invention provides a multi-party record linkage method, comprising the steps of: data preprocessing: dividing data sources of a plurality of participants into blocks, and converting records in thedata sources into bit arrays; record approximate matching: calculating the ratio of the bit arrays corresponding to the position of bit 1, and determining the corresponding positions of the interestrate as candidate matching positions when the interest rate reaches a dynamic threshold; similarity calculation: calculating the similarity between the candidate matching positions and judging whetherthe similarity reaches the global threshold. If yes, the matching is successful, otherwise the matching fails. The invention also relates to an electronic device, a storage medium and a multi-party record linkage system. The invention adopts the ratio to mark the similarity between the records at a certain position, thereby improving the fault tolerance; By using dynamic threshold and candidatematching positions to check and determine the successful matching position, so that the recall ratio and precision ratio are high; the similarity between records with quality problems can be effectively calculated; the invention realizes multi-party record linkage, can effectively protect privacy, and has good scalability and fault tolerance.

Description

technical field [0001] The invention relates to the technical field of record linking, in particular to a multi-party record linking method, electronic equipment, storage medium and system. Background technique [0002] With the continuous advancement of technology, data is growing and accumulating rapidly, and the era of big data has arrived. How to organize and analyze these data is the key to realizing the value of data. However, the existing record linking methods do not consider the privacy protection of recorded information when the recorded information involves personal privacy or sensitive information. Existing record linking methods are only applicable to two data sources, but many applications in reality often have more than two data sources for record linking. With the ever-increasing amount of data and the existence of real-world data quality problems, such as spelling mistakes, order reversal, etc., existing record linking methods are poor in scalability and f...

AI Technical Summary

Problems solved by technology

[0003] In order to overcome the deficiencies of the prior art, one of the objectives of the present invention is to provide a multi-party record linking method, which solve...

Images