Personalized calibration set selecting and modeling method for spectrum sample
A modeling method and a correction set technology, applied in the field of near-infrared spectral analysis, can solve the problems of not considering the spectral information of unknown samples, and the difficulty of determining the prediction of unknown samples, so as to achieve the effects of saving manpower and material resources, good prediction effect, and precise distribution
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0057] Taking the public corn data as an example, there are a total of 80 samples measured, including the repetition of samples. Matrix X is the original near-infrared spectrum matrix of corn samples, and matrix Y is the matrix of four quality index components (water, oil, protein, starch).
[0058] For the four components of the Y matrix, each column in the matrix Y is associated with the matrix X to model respectively. In this embodiment, water is used as an example to illustrate the method, and the same steps are taken for the remaining components and water. For the original spectrum of the sample, see figure 1 shown.
[0059] Firstly, the abnormal samples are eliminated, through Hotelling T 2 The method detects the original spectral matrix X, and obtains 3 abnormal samples, and detects the reference value matrix Y through the Boxplot method, and there are no abnormal values, and the remaining 77 samples are removed, and the X m matrix. The principal component projection...
Embodiment 2
[0088] Taking the public data corn as an example, there are 80 samples tested. X is the near-infrared spectrum matrix of the sample, and Y is the matrix of four component quality indicators. Take water as the object description, and take the same steps for the rest of the ingredients, first remove the abnormal samples, and use Hotelling T 2 method, 3 abnormal samples were detected, and a total of 77 samples were left after elimination. We changed the way of calculating the distance to investigate the impact of various division methods on the performance of the model after the way of calculating the distance was changed.
[0089] Randomly draw 10 samples as an independent validation set X t .
[0090] Divide the remaining 67 samples and calculate X t Each sample and the remaining samples X in k Mahalanobis distance between D tk And sorted, for the independent validation set X t Select the most similar sample (ie g=1) for each sample to form the final verification set X v...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More