| 학회 |
한국공업화학회 |
| 학술대회 |
2020년 가을 (10/28 ~ 10/30, 광주 김대중컨벤션센터(Kimdaejung Convention Center)) |
| 권호 |
24권 1호 |
| 발표분야 |
학생우수논문발표(석사과정) |
| 제목 |
Effect of data pre-processing on the performance and robustness of prediction model for distillation column |
| 초록 |
Though the performance of the predictive model depends on how accurate and how much information data have, the raw process data has low information by including noise, outliers, and defects. Therefore, it is necessary to conduct appropriate data pre-processing to improve the performance of the precess data-driven predictive model. In this study, we suggested two outlier detection and one noise filtering for data pre-processing and applied each method with various ranges on the raw process data of the distillation column. Then we evaluated the effect of the pre-processing on the LSTM-based model performance. The statistical methods were used for performance evaluation, and each predictive model was conducted thirty times to confirm the performance and robustness. As an effect on applying the data pre-processing, we found that the model showed more robust prediction results, and the accuracy and precision were increased by 26.16% and 5.26%, respectively. |
| 저자 |
최영렬1, 권혁원1, 이예찬1, 박현도1, 문일2, 조형태1, 김정환1
|
| 소속 |
1한국생산기술(연), 2연세대 |
| 키워드 |
Data pre-processing; LSTM; predictive model; distillation
|
| E-Mail |
|
