0

I don't have much of a background in coding.

I have a point layer with 400 points. The attribute table contains 3 fields of interest:

  • Bearing (values between 0-180);
  • a Ratio (ratio of ellipse long/short axes - continuous values between 0-8);
  • text descriptor (finite range of text values [eg. could be same descriptor for n points or 0 points])

I want to somehow rank the records in group 1 according to each record's similarity to all other records as well as the points' (2D) distance between each other. The goal is to use a procedural method to create relative values for each point that can be used to point towards potential groups of points.

However, I want to place the highest weight on the 'text descriptor' field and second-highest on proximity, because these will be the fundamental indicators of similarity. The other 2 fields ('ratio', 'bearing') will serve to increase the likelihood of group membership after the first two have been evaluated and these latter 2 are of the lowest and equal weighting to each other.

From what I've read, it seems like hierarchical clustering could work for this kind of problem. Maybe AHP would work? Could anyone explain how I could apply the right analysis to this problem, preferably within QGIS?

2
  • Please provide a reproducible example representing your data and what you have tried to date. How exactly do you expect clustering of the "text descriptor" to work? Especially in regard to providing "weight" to the other parameters? You are delving into statistical topics such as NLP and not spatial analysis per se, or even vanilla clustering. Using common clustering methods, you cannot derive any type of distance measure, used to identify clusters, with a text field. If you can figure out some type of ranks for your text you will be in much better shape. QGIS will not be of much help here. Commented Sep 7, 2021 at 17:03
  • Hearing that it seems like this is probably the wrong place to post this. Also, I'm definitely in over my head. Ill try editing my question once to make the problem clearer but if that doesn't work I'll delete it and try somewhere else. Thank you for your input!
    – GeoStudent
    Commented Sep 8, 2021 at 18:56

0

Your Answer

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

Browse other questions tagged or ask your own question.