Using Neural Networks for Addressing Data Quality During the Software Maintenance Process

摘要

The high cost of software maintenance continues to he a great concent for many organizations due to poor data quality that plagues most legacy database systems. It is proposed in this paper that neural net technology be used to accommodate changes inuser requirements when data quality is an issue. Neural nets can be trained to identify semantically equivalent data such that source code modifications do not have to be made. A case study is used to illustrate the use of neural nets to replace source code in identifying duplicate data within and across databases even when data is incorrect or incomplete.