Algorithm for detecting approximate duplicate records in massive data