| Author |
 Topic  |
|
|
kubra
Slovakia
10 Posts |
 Posted - 11/02/2015 : 09:16:45 AM
|
Origin Ver. and Service Release (Select Help-->About Origin): 8/8.5
Operating System:Win7
Hello
I have an array of many data points (it depends on measurment, but around few thousands). These data points are not regularly distributed in array, because of measurment errors. However, i need to go through measured points (one by one), find their nearest neighbours and store x and y values in vector. Number of nearest neighbours is given by constant (between 25 and 121, from SF value given by (2*SF+1)^2). I know, these numbers are huge, and i am ok with long duration of calculating.
My algorithm does that. However, it is written to find at least so many nearest neighbours, as the constant is set. So, if constant is 25, for various points it finds 27, 28, 30 etc. points, and stops, when number of them exceed value of constant. If it is less, than value of constant, it erases all values in vector, rises up array, in which it is searching for NN, and does that again. Therefore calculation time (which is by itself long enough) is extending.
My code:
for(C=1; C<=nmbCols-1; C++){
for(R=0; R<=nmbRows-1; R++){
if(!is_missing_value(wks.Cell(R,C))){
i=SF;
j=SF;
do{
for(rR=R-j; rR<=R+j; rR++){
if(rR<0 || rR > nmbRows-1){
;
}
if(rR>=0){
for(cC=C-i; cC<=C+i; cC++){
if(cC<=0 || cC > nmbCols-1){
;
}
if(cC>0){
if(!is_missing_value(wks.Cell(rR,cC))){
vecH.Add(wks.Cell(rR, 0));
vecM.Add(wks.Cell(rR, cC));
}
if(is_missing_value(wks.Cell(rR,cC))){
;
}
if(cC==C+i){
break;
}
}
}
}
if(rR==R+j){
break;
}
}
if((vecM.GetSize())<((2*SF+1)^2)){
i++;
j++;
vecH.RemoveAll();
vecM.RemoveAll();
}
if((vecM.GetSize())>=((2*SF+1)^2)){
;
}
}while((vecM.GetSize())<((2*SF+1)^2));
After code is finished, program does euclidian distances calculations from the calculated data point, and sorts them in vector, and by that it chooses desired number of NN.
I know, it is very specific problem. Maybe if somebody has already done NN algorithm, or there is Origin C command i haven´t found...or has an idea for such edit of code, so it would not have to do repeating calculations... I would really appreciate it :)
|
Edited by - kubra on 11/02/2015 09:23:17 AM |
|
|
SeanMao
China
288 Posts |
Posted - 11/03/2015 : 05:28:10 AM
|
Hi,
We have a built-in X-function levelcrossing that is supported since Origin 8.1 that may help you find the NN X coordinates at a given level. You can find the reference from the link below:
http://www.originlab.com/doc/X-Function/ref/levelcrossing
Give it a try and see if could be some help.
Regards!
Sean
OriginLab Tech. Service |
 |
|
|
kubra
Slovakia
10 Posts |
Posted - 11/03/2015 : 1:10:18 PM
|
Hello Sean
Thank you for reply. However, data in my dataset are more-less randomly distributed, and i can´t say beforehand how big is the region, where NN will be. Also, minimum required version is 8.1 SR0 (8.6, eventually), and i need to run my program on version 8.
But thanks again for reply!
Have a nice day
Branislav |
|
|
| |
Topic |
|
|
|
nearest neighbours algorithm