IJPEM-ST

Integrating Entropy-based Data Reduction and Machine Learning in Multidisciplinary Engineering Systems for Enhanced Response Prediction
A New Approach to Improving System Response Prediction: Combining Entropy-Based Mining and Machine Learning


Seung-Kyum Choi/Georgia Institute of Technology


  • Keywords: Data Uncertainty (redundancy), Data Mining, Entropy, Machine learning, Surrogate Modeling
  1. The study introduces a framework for modeling streamlined design variables in multidisciplinary engineering systems. These systems often deal with uncertainties, and correlated design variables can lead to data redundancy, which affects system response predictions.
  2. To address this issue, the framework employs data reduction techniques based on the entropy-based correlation coefficient (referred to as “e”). There are two distinct approaches:
    • For strong correlations (high e values), feature extraction is used.
    • For weak correlations (low e values), feature selection is employed.
  3. Additionally, the study integrates an Artificial Neural Network (ANN) to efficiently predict complex responses.
  4. The framework’s efficacy is demonstrated through Multiphysics applications such as an electro-mechanical stretchable patch antenna. The antenna’s performance is influenced by the roughness of its patch surface, which exhibits a strong correlation.
 
논문 무료 다운로드
[Free Access Period: 24. 8. 3. ~ 24. 9. 30.]