Data synthesis method preserving correlation of features

作者：

Highlights：

• The previous data synthesis method cannot preserve the correlation between features of an original dataset.

• A new data synthesis method was developed to create high quality data by maintaining the original correlations.

• The new method can rapidly create artificial datasets by using linear algebra techniques.

• The results showed that the classification accuracy was significantly improved by using the artificial dataset.

摘要

•The previous data synthesis method cannot preserve the correlation between features of an original dataset.•A new data synthesis method was developed to create high quality data by maintaining the original correlations.•The new method can rapidly create artificial datasets by using linear algebra techniques.•The results showed that the classification accuracy was significantly improved by using the artificial dataset.

论文关键词：Data synthesis,Correlation,Artificial dataset,Random noise

论文评审过程：Received 30 July 2020, Revised 20 May 2021, Accepted 8 August 2021, Available online 26 August 2021, Version of Record 3 September 2021.

论文官网地址：https://doi.org/10.1016/j.patcog.2021.108241