Normalize data before measuring the distance

2 次查看（过去 30 天）
KnowledgeSeeker
KnowledgeSeeker 2014-3-20
回答： Shashank Prasanna 2014-3-20
I am trying to measure the correlation between data (a row vector). I use euclidean distance to measure how correlated each feature to one another. I want to measure the distance of the first feature against the rest of the features. First I normalize my data:
normalizedData = x/sum(X)
second I use Euclidean distance to measure the distance of one feature to the rest of the features
mydata = {}
for Idx = 1:numel(normalizedData)
euclidean_distance = normalizedData(1) - normailizedData(Idx)
mydata{1,1} = euclidean_distance;
end
I am wondering if the above way is correct or incorrect? can someone suggest me or give me an opinion.
Thank you

请先登录，再回答此问题。

回答（1 个）

Shashank Prasanna
Shashank Prasanna 2014-3-20
You can use pdist2 to compute pairwise distances:
http://www.mathworks.com/help/stats/pdist2.html
If you want to standardize the data before hand, you can use zscore:
http://www.mathworks.com/help/stats/zscore.html
But if you want to find correlation between to features you use corr
corr(X(:,1),X(:,2))
http://www.mathworks.com/help/stats/corr.html

类别

在 Help Center 和 File Exchange 中查找有关 Measurements and Feature Extraction 的更多信息

标签

产品

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by