Knn score distance, Probablity density function, Histogram

Question

Natalie 2013-6-14

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/79133-knn-score-distance-probablity-density-function-histogram

I am presently running a knnscoredistance test using two sets of data- a test and model set. I have then normalized the data using the norm command and plotted it as a histogram. However, I need to plot the data with the probability density function taken to really make the information meaningful. I have been unsuccessful at this for many days now and was curious if there is a method or command to do this. Also I think it could be interesting to run this test using various K values (1-10 at first) to generate 10 separate histograms but in the same plot. Do you know if this is an option. As of now I am doing this rewriting the code with the new value, but is a more concise method id available that would be ideal. I sincerely appreciate your time and help.

% INPUTS:
%   xref  = dataset of Training Set in evrimodel=.
%   xtest = a dataset of testset in evrimodel.
xtest=importdata('Model_TestSet-Fixed.mat');
xref= importdata('Model_TrainingSet.mat');
%k=number of components
k=3;
%%creates a predicitive model based off training and test set 
model=[knnscoredistance(xref,k)]
model_2=[knnscoredistance(xtest,k)]
model=model/norm(model)%%normalize data by plotting frequency on Y-axis
model_2=model_2/norm(model_2)
%%create histogram comparing both models
clf %clear figure window
hist(model,18)%create histogram with (data,nbins)
hold on
hist(model_2,18)
%%Training set in red and Test set in Green
h = findobj(gca,'Type','patch');
display(h) 
set(h(1),'FaceColor','r','EdgeColor','k');
set(h(2),'FaceColor','g','EdgeColor','k');
set(gca,'xscale','log');%%log of x-axis
%%Labeling the histogram
xlabel('Standardized Distance (log)'); 
ylabel('#');
title('Knn Score Distance of Training and Test Set');
legend('Training Set','Test Set');
figure(1)