Duplicate Data Detection Using GNN

In some applications like data warehousing, data mining or information integration data must be cleaned as a preprocessing step to ensure the quality of data and the performance of applications. An essential work in data cleaning is duplicate record detection. Existing detection methods apply to different data models and record types. Certain difficulties with those studies are still to be overcome. This paper proposes a genetic neural network based approach to duplicate record detection. The topology and weight vector of a neural network are firstly optimized by a genetic algorithm for the given data set before it is used to perform the detection. The method can enhance the detection accuracy and alleviate the many problems with previous works.