Finding k-nearest neighbors to get outliers?
2 views (last 30 days)
Show older comments
i want to implement k-nearest neighbors to get outliers with better time complexity
steps:
1-skip first distance for knn(equal zero).
2-Get mean of every distance point in query .
3-Decceding all mean- distance points to get top n (20) .
4- Get size of real outlier from top n to calculate outlier detection rate.
[Idx,Dist] = knnsearch(Data,query,'k',10,'Distance','cityblock','NSMethod','kdtree','IncludeTies',true);
t=mean(Dist(:,2:end),2);
[distance, index] = sort( t, 'descend');
out=(query(index(1:20, 1),:));
[~,ia,~] = intersect(datasetoutlier(:,:),query(:,:),'rows');
nreal-outlier=size(ia);
nout-dataset = size(datasetout(:,:), 1);
detection_rate=nreal-outlier/nout-dataset
1- I want to know if my code is correct to use for big dataset ?.
I found this code in matlab center but i don't know much about parfor .
2-Does this code will be better performance(run time) because of parfor?
function [id,dist] = matlabKNN(data,k,disMetric)
if strcmp(disMetric,'euclidean')
disMetric = 'euclidean';
else
error("δ֪µÄ¾àÀë²ÎÊý");
end
k = k + 1;
[n,~] = size(data);
id = zeros(n,k);
dist = zeros(n,k);
parfor i = 1:1:n
datum = data(i,:);
[aID,aDist] = knnsearch(data,datum,'K',k,...
'Distance',disMetric,'IncludeTies',true);
aID = aID{1}(1:(k));
aDist = aDist{1}(1:(k));
id(i,:) = aID;
dist(i,:) = aDist;
end
id1 = id(:,1);
dist1 = dist(:,1);
id = id(:,2:end);
dist = dist(:,2:end);
for i = 1:1:n
mySelf = find(id(i,:) == i);
if(~isempty(mySelf))
id(i,mySelf) = id1(i);
dist(i,mySelf) = dist1(i);
end
end
end
thanks
0 Comments
Answers (0)
See Also
Categories
Find more on Statistics and Machine Learning Toolbox in Help Center and File Exchange
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!