Semi-supervised chemometric methodology for improved multivariate calibration of spectrometers.
In this work, a semi-supervised chemometric methodology is designed to improve the predicted accuracy for the multivariate calibration in spectrometers. The simulation is carried out on MATLAB platform. The available toolbox of partial least-squares regression (PLS), which is a very powerful algorit...
Saved in:
| 主要作者: | |
|---|---|
| 其他作者: | |
| 格式: | Final Year Project |
| 語言: | English |
| 出版: |
2010
|
| 主題: | |
| 在線閱讀: | http://hdl.handle.net/10356/39522 |
| 標簽: |
添加標簽
沒有標簽, 成為第一個標記此記錄!
|
| 總結: | In this work, a semi-supervised chemometric methodology is designed to improve the predicted accuracy for the multivariate calibration in spectrometers. The simulation is carried out on MATLAB platform. The available toolbox of partial least-squares regression (PLS), which is a very powerful algorithm designed to build quantitative models, is used extensively in this work. Three algorithms are developed and evaluated based on the given data set. The finalized algorithm, which utilizes the co-training method, is further evaluated with three different data sets. The results of the prediction accuracy improvement are obtained and analyzed. Although the improvement is not significant, the feasibility of the algorithm is still valuable to be discussed. Recommendations on future directions are given to the development of a better algorithm. |
|---|