Jump to content



Photo
- - - - -

Hierarchical Clustering [Fx-5800P]

fx-5800P

  • Please log in to reply
No replies to this topic

#1 Tritonio

Tritonio

    Newbie

  • Members
  • Pip
  • 15 posts
  • Gender:Male

  • Calculators:
    FX-5800P, FX-991ES+

Posted 13 November 2014 - 03:11 PM

Here's another clustering algorithm for data points in the stats table. Like K Means it doesn't read data from the frequency column so multiplicities have to be types in multiple rows instead of using the Freq column. The freq column will contain the number of the cluster that each point is assigned at the end.
 
The distance function used is the euclidean distance of the centers of the clusters so this is a bad (but fast) hierarchical clustering algorithm.
 
In the future I will probably update it to support single-linkage distance but it will be REALLY slow.

FreqOff
n→T
FreqOn
0→DimZ
3T+3→DimZ
For 1→I To T
List X[I]→Z[3I]
List Y[I]→Z[3I+1]
1→Z[3I+2]
I→List Freq[I]
Next
"CLUSTERS"?C
For T→A To C+1 Step -1
10×10^99→B
For 1→I To A-1
For I+1→J To A
√((Z[3I]-Z[3J])²+(Z[3I+1]-Z[3J+1])²)→D
If D<B:Then
D→B
I→V
J→W
IfEnd
Next
Next
List Freq[W]→List Freq[V]
(Z[3V]×Z[3V+2]+Z[3W]×Z[3W+2])÷(Z[3V+2]+Z[3W+2])→Z[3V]
(Z[3V+1]×Z[3V+2]+Z[3W+1]×Z[3W+2])÷(Z[3V+2]+Z[3W+2])→Z[3V+1]
Z[3V+2]+Z[3W+2]→Z[3V+2]
Z[3A]→Z[3W]
Z[3A+1]→Z[3W+1]
Z[3A+2]→Z[3W+2]
Cls
Locate 1,1,"CLUSTERS"
Locate 10,1,A-1
Next
Cls
Locate 1,1,"ASSIGNING"
For 1→I To T
10×10^90→B
For 1→Q To C
√((List X[I]-Z[3Q])²+(List Y[I]-Z[3Q+1])²)→D
If D<B:Then
D→B
Q→W
IfEnd
Next
W→List Freq[I]
Next
Cls
"DONE"






Also tagged with one or more of these keywords: fx-5800P

0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users