Insufficient number of samples (Feature engineering)


After applying function of abun_per_tempbin (freq=100), I noticed that the number of samples was insufficient.

Let’s look at the graphics.
Selected values ​​to be shown as examples:

data: “S0000”
m/z = 18
freq = 100
n_samples = 193 → 16

More accurate feature can be obtained with slightly more (25) than 16 samples.
Let’s look at the graphics.

if you increase number of scaling(0,100) abundance(zooming the peaks), you can decrease number of samples. So you can preserve the shape of the peaks including height, width, and area
Onur Koç.