Solving problems while combining data between different industries through sequential data combination process
Sumin Jeong1 · Hyeonjin Oh2 · Eunhye Choung3 · Suhyun Cho4
1Department of Statistics, Kyungpook National University
2Department of Software, Korea National University of Transportation
3Department of Psychology, Sungshin Women’s University
4Department of AI Information Science, Myongji University
2Department of Software, Korea National University of Transportation
3Department of Psychology, Sungshin Women’s University
4Department of AI Information Science, Myongji University
Correspondence to: 1 Bachelor’s graduate, Department of Statistics, Daegu 41566, Korea.
2 Bachelor’s graduate, Department of Software, Chungju 27469, Korea.
3 Master’s graduate, Department of Psychology, Seoul 02844, Korea.
4 Professor, Department of AI Information Science, Myongji University, Seoul 03674, Korea. E-mail: whtngus3232@gmail.com
2 Bachelor’s graduate, Department of Software, Chungju 27469, Korea.
3 Master’s graduate, Department of Psychology, Seoul 02844, Korea.
4 Professor, Department of AI Information Science, Myongji University, Seoul 03674, Korea. E-mail: whtngus3232@gmail.com
Received May 12, 2024; Revised July 1, 2024; Accepted August 6, 2024.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/3.0) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.
- Abstract
- As the importance of personalized services increases, data combination across different industries is becoming more active. However, existing data combining methods have the problem of high possibility of personal information leakage because of the direct exchange of data containing sensitive information. So in this study, we suggest a data combination method using minimal unique information such as gender and age. We used Federated Learning and Split Learning to enhance data security when training the model. As a result, combining data through the SEC process showed improved prediction performance compared to raw data. Data combination using minimal unique information through the SEC process enhances privacy protection by preventing infringement of personal information. So we expect this method can not only improving problems with existing data combination methods, but also be applied to various services through improved predictions.
- Keywords : Data combination across different industries, Federated learning, Privacy protection, SEC process, SHAP, Split learning, STC process
Full Text(PDF) Free
