Feedback Clustering Algorithm for Detecting Approximately Duplicate Records
Dong Mei Hu, Xiao Feng Zhou, Xiao Feng Chen · Applied Mechanics and Materials · 2014
Detecting and merging approximately duplicate records is not an emerging issue in the field of data cleansing, the majority of duplicated records detecting method is based on the "sort-merge" thinking. Although clustering methods have been applied to data cleaning, a large number of non-duplicated records exist in clusters after analysis as a result of the increasing records. Response to this shortcoming, this paper presents a data cleansing method based on Clustering Feedback Pattern. Comparison results of clustering are fed back to the cluster process so that recall and precision improve.