How I organize and insert NIR dataset for Classification Learner.

Hello,

I currently research the field of near-infrared spectroscopy (NIR), mainly in the analysis of the spectra of polymeric materials. I would like to make a system for classifying these spectra by machine learning. Literature indicates the APP Classification Leaner and its models to perform the training of the spectra. However, I'm having trouble inserting the spectra sheet (containing wavelength and absorbance values). Being new to Matlab, I don't know how to organize these spreadsheets with different values, because the wavelength is constant for the different samples, but the absorbance varies for them. My idea is to classify the spectra for the different materials. How could I organize the spreadsheet and develop the training?

If anyone can help me with a tutorial or quick step by step, I would be extremely grateful.

Thanks.

[KHAN WALI]

Hi,

I think your question is old and you already got the answer. I just started dealing with spectral analysis so here is my answer.

first you have some spectral data say m sample in the following order

Sample# wavelegth1 wavelength 2 ....................wavelenght N

1 0.89 0.7 ................................0.1

2 0.7 0.78 .................................0.9

3 . . ................................. .

. . . ................................... .

.

.

m 0.9 0.6 ...................................0.9

remember wavelength 1 means say 450, 0.89, 0.7 etc are corresponding abosorption/reflectance etc.

so you have n*m matrix.

now what you want to look for? is it some kind of material and you want to find some particular element? if yes the just go and od some chamical analysis (wet chemistry) tests and find the exact amount of that element in each sample. store it in a different column.

say we are looking for element "X" content in all the sample, so it will for a column vector (m*1)

you are ready to apply any method say deep learning for classification.

use some of the data as a training set, validation and testing.

hope it help

Best Regards

[Godwin Jr.]

I tried this method but all models work great on training sets but are terrible on test data. Any help with machine learning workflows for NIR Spectroscopy in MATLAB?

[jianjun]

Hello. I looked this interesting comments. I have another question, that is, how can I use matlab+SOLO APP to give a classification analysis without Y value when establishing NIR model with only wavalength values? For example, I have 200 sample, these samples are individual grain kernel, and I want to see if they can be grouped by their weavelength information? If yes, how to realize it? Thanks!