machine learning - Clustering algorithm for gps data -
i have data set consisting of gps coordinates points on particular city (let's take san francisco example). want cluster coordinates groups such in image:
should use k-means or dbscan or other clustering algorithm? should find clusters first , find border points draw boundaries?
the example partitioning showed looks more forced quantization me clustering based on structure in data set.
which algorithm choose depends on
- your data (not data type, actual distribution of values have)
- your needs (what want get, algorithms built based on different desires)
but don't have data, , don't know desires. except sketch did looks more k-means quantization density based clustering. except k-means minimizes variance, , can't handle latitude, longitude well. can project data appropriate utm zone though; k-means should work.
Comments
Post a Comment