0
$\begingroup$

I am given a dataset where I have to predict the distance. The training dataset consists of the values of the signal strength of 5 different sensors and the distance between the source and the 4 points where those 5 sensors are placed.

Example

enter image description here

Based on this kind of training dataset, I have to predict distance for a particular vector of signal strength. Here D1, D2 are numerical values.

Can one guide which regressor model and what preprocessing technique will be useful here?

I am new to the ML domain!

Improve this question
$\endgroup$
6
  • $\begingroup$ From the table it looks like the 'distance' is not numerical but categorical: D1, D2,D3, D4. If so this is not regression but classification. Additionally does your dataset have only 4 instances? If yes this is way too small for proper statistical learning. You could just calculate a distance with D1, D2, D3,D4 based on the features (sensors) and then pick the min. It would be equivalent to using kNN with k=1. $\endgroup$
    – Erwan
    Commented Oct 11, 2022 at 21:56
  • $\begingroup$ @Erwan D's are numerical, I have mentioned. This example here is just a sample. There are all total 1000+ such instances. $\endgroup$
    – XYZ
    Commented Oct 12, 2022 at 3:07
  • $\begingroup$ So all the Ds are different, they don't correspond to a fixed set of locations as indicated in this table? $\endgroup$
    – Erwan
    Commented Oct 12, 2022 at 9:36
  • $\begingroup$ D1 is the distance between P1 and Source; D1 = EucliD(Source,P1). Rest are also same! $\endgroup$
    – XYZ
    Commented Oct 12, 2022 at 9:54
  • $\begingroup$ If the locations are all different with each other, then you're correct that this is a regression task. You have many choices, I often recommend decision trees or SVM regression. Are the values of the sensors normalized? The ones in the table look like [0-100], but if not they should be scaled for SVM. $\endgroup$
    – Erwan
    Commented Oct 12, 2022 at 11:03

0

Reset to default

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.