Clustering GPS data using DBSCAN, but clusters don't make sense (in terms of size)

Question

I work with GPS data (latitude, longitude). For density-based clustering, I used DBSCAN in R.

Advantages of DBSCAN in my case:

• I do not need to predetermine the number of clusters

• I can calculate the distance matrix (using the Haversine Distance Formula) and use this as an input to dbscan

  library(fossil)
  dist<- earth.dist(df, dist=T) #df is dataset containing lat long values
  library(fpc)
  dens<-dbscan(dist,MinPts=25,eps=0.43,method="dist")
  

Now, when I look at clusters, they do not make sense. Some clusters have points located more than 1 km apart. I want dense clusters, but not so big.

Various values MinPtsand eps are served , and I also used the k nearest neighboring graph to get the optimal value epsfor MinPts= 25

, dbscan, , p MinPts eps, , , (, , ).

, " , ", :

• ? , dens$cluster, , - ​​ ?
• 0?
• , eps. , .
• - , dbscan ?

OPTICS - , ?

Note: , . , 1 , .

Answer 1

DBSCAN , .

? ; eps - ; , ; .

. , , .

, 0 R. R, waaaay , . R, , ELKI. DBSCAN ELKI, LatLngDistanceFunction R- . , , R.

. , - , ?

, ( , )

, .