Enhanced Taguchi's T-method using angle modulated Bat algorithm for prediction
Analysis of multivariate historical information in predicting future state or unknown outcomes is the core function of Taguchi’s T-method. Introduced by Dr. Genichi Taguchi under Mahalanobis-Taguchi system, the T-method combines regression principle and robust quality engineering element in formulat...
保存先:
| 主要な著者: | , , , |
|---|---|
| フォーマット: | 論文 |
| 言語: | English |
| 出版事項: |
Institute of Advanced Engineering and Science (IAES)
2022
|
| 主題: | |
| オンライン･アクセス: | http://eprints.utm.my/id/eprint/98611/1/FaizirRamlie2022_EnhancedTaguchisTMethodusingAngleModulated.pdf http://eprints.utm.my/id/eprint/98611/ http://dx.doi.org/10.11591/eei.v11i5.4350 |
| タグ: |
タグ追加
タグなし, このレコードへの初めてのタグを付けませんか!
|
| 要約: | Analysis of multivariate historical information in predicting future state or unknown outcomes is the core function of Taguchi’s T-method. Introduced by Dr. Genichi Taguchi under Mahalanobis-Taguchi system, the T-method combines regression principle and robust quality engineering element in formulating a predictive model and employs taguchi’s orthogonal array design in optimizing the model through feature or variable selection process. There is a concern regarding the sub-optimality of the T-method prediction accuracy, particularly when the orthogonal array failed to offer a significant number of combinations in search for an optimal subset of features. This is due to the fixed and limited combination offered for evaluation as well as the lack of higher-order interaction of combination. In response to this issue, this paper proposed an angle modulated Bat algorithm to be integrated with the T-method in optimizing the prediction model. A comparison study was conducted using energy efficiency benchmark datasets with the mean absolute error metric used as the performance measure. The results show that the proposed method improved the prediction accuracy by 10.74%, from 6.05 to 5.4, by integrating only four features over the original eight in the prediction model. |
|---|