bloods.ai Blood Spectroscopy Classification Challenge
Helping Global
$7 500 USD
Completed (~4 years ago)
Classification
1103 joined
265 active
Start
Oct 19, 21
Close
Feb 13, 22
Reveal
Feb 13, 22
Clarification on Zindi Contest Spectra
Data · 18 Dec 2021, 19:11 · 5
Hello,
Please in Zindi Contest Spectra dataset, Glucose, Cholesterol and skinbloodfat have absorbance values up to 170 while Hemoglobin has absorbance up to 148.
I'm assuming 170 corresponds similar to 170 absorbance values in train and 148 absorbance values matches train trimmed size wavelength range.
Would it be valid to remove first eleven and last eleven for Glucose, Cholesterol and skinbloodfat to have the trimmed wavelength range of 148 absorbance values too?
Thank you.
If trimming the wavelength range improves your out of fold cross-validation scores, then I would think that is an example of effective feature selection. I have experimented with different wavelength ranges and found that trimming the edges had a very small effect on the validation error.
Don't let that discourage you from experimenting with different wavelength windows though
Thank you for your response. I just wanted to be sure trimming first and last eleven in Zindi Spectra data to obtain a trimmed wavelength range is equivalent to train dataset trimmed range. In a previous discussion, it was suggested trimming train this way gives train trimmed.
Hi (-:
The Hemoglobin grapg start from 900nm (our data wavelength range is 900nm - 1700nm) and ends at ~1595.8 - so you have to trim the last absorbance items of the data to get 148 array
Yes thank you for the response Maya.
Interesting observation about the Zindi Contest Spectra data! It sounds like you're really digging into the details. Aligning the absorbance ranges by trimming Glucose, Cholesterol, and skinbloodfat could be a valid approach to standardize the data. Speaking of alignment, sometimes I feel like I'm navigating tricky terrain just like in Snow Rider 3D. Precision is key both in data analysis and on those virtual snowy slopes!