Метод k взвешенных ближайших соседей (пример)

%% Example 1
% Example, which is simply classifed by WeightedKNN. It seems like XOR
% problem.
 
%% generate training sample
% generate 2 groups of normal classes. Each group consistes of 2 simple
% normal classes
XL1 = [GetNormClass(50,[0,0],[1,1]); GetNormClass(50,[6,6],[1,1])];
XL2 = [GetNormClass(50,[6,0],[1,1]); GetNormClass(50,[0,6],[1,1])]; 
XL = [XL1; XL2];
YL = [repmat(1,100,1);repmat(2,100,1)];
 
%% generate control data with the same distribution
X1 = [GetNormClass(50,[0,0],[1,1]); GetNormClass(50,[6,6],[1,1])];
X2 = [GetNormClass(50,[6,0],[1,1]); GetNormClass(50,[0,6],[1,1])];
X = [X1; X2];
Y = [repmat(1,100,1);repmat(2,100,1)];
 
%% get classification
%% choosing parametrs
PP = ParAdjust(XL, YL);
PP.XL = XL;
PP.YL = YL;
%% classification
Y = WeightKNN(X, PP);
 
%% results visuaisation
%% plotting real classes of objects
plot(X1(:,1),X1(:,2),'*r');
hold on
plot(X2(:,1),X2(:,2),'*b');
 
%% plotting classification results
plot(X(Y == 1,1),X(Y == 1,2),'or');
plot(X(Y == 2,1),X(Y == 2,2),'ob');
hold off

%% Example 2
% Classes cross each other in this example. Algorithm makes misstakes in
% classification.
 
%% generate training sample
% generating 4 sample normal classes
XL1 = GetNormClass(100,[5,0],[10,2]); 
XL2 = GetNormClass(100,[5,10],[10,2]);
XL3 = GetNormClass(100,[0,5],[4,4]);
XL4 = GetNormClass(100,[10,5],[4,4]);
 
XL = [XL1; XL2; XL3; XL4];
YL = [repmat(1,100,1);repmat(2,100,1);repmat(3,100,1);repmat(4,100,1)];
 
%% generate control data with the same distribution
X1 = GetNormClass(100,[5,0],[10,2]); 
X2 = GetNormClass(100,[5,10],[10,2]);
X3 = GetNormClass(100,[0,5],[4,4]);
X4 = GetNormClass(100,[10,5],[4,4]);
X = [X1; X2; X3; X4];
% X is going to be changed in getting classification. X0 is needed to plot data.  
X0 = X;
 
%% getting classification
%% features standardization
 [p, m] = size(X);
 [n, m] = size(XL);
 Z = [XL; X];
 Z =FeaturesStand(Z);
 XL = Z(1:n, :); 
 X = Z(n+1:n+p, :);
%% choosing parametrs
PP = ParAdjust(XL, YL);
PP.XL = XL;
PP.YL = YL;
%% classification
Y = WeightKNN(X, PP);
 
%% results visuaisation
%% plotting real classes of objects
plot(X1(:,1),X1(:,2),'*r');
hold on
plot(X2(:,1),X2(:,2),'*b');
plot(X3(:,1),X3(:,2),'*g');
plot(X4(:,1),X4(:,2),'*y');
%% plotting classification results
plot(X0(Y == 1,1),X0(Y == 1,2),'or');
plot(X0(Y == 2,1),X0(Y == 2,2),'ob');
plot(X0(Y == 3,1),X0(Y == 3,2),'og');
plot(X0(Y == 4,1),X0(Y == 4,2),'oy');
%% count errors
errors = sum([Y(1:100) == 1; Y(101:200) == 2; Y(201:300) == 3; Y(301:400) == 4])
 
hold off

%% Example 'iris'
% Example with real data
 
%% data preparation
%% load data
load 'iris.txt';
S = iris;
%% eliminating first two attributes
S(:,1:2) = [];
%% divide data into training part and pert to classify
XL = S([1:25,51:75,101:125],:);
X = S([26:50,76:100,126:150],:);
%% creating class labels
YL = [ones([25,1]);2*ones([25,1]);3*ones([25,1])]; 
Y = [ones([25,1]);2*ones([25,1]);3*ones([25,1])];
 
%% plotting real classes of objects
plot(X(Y == 1,1),X(Y == 1,2),'*r');
hold on
plot(X(Y == 2,1),X(Y == 2,2),'*b');
plot(X(Y == 3,1),X(Y == 3,2),'*g');
 
%% getting classification
%% features standardization
 [p, m] = size(X);
 [n, m] = size(XL);
 Z = [XL; X];
 Z =FeaturesStand(Z);
 XL = Z(1:n, :); 
 X = Z(n+1:n+p, :);
%% choosing parametrs
PP = ParAdjust(XL, YL);
PP.XL = XL;
PP.YL = YL;
%% classification
Y = WeightKNN(X, PP);
 
%% plotting resulting classification
plot(X(Y == 1,1),X(Y == 1,2),'or');
plot(X(Y == 2,1),X(Y == 2,2),'ob');
plot(X(Y == 3,1),X(Y == 3,2),'og');
title('Irises classification')
xlabel('petal width, cm');
ylabel('petal length, cm');
legend('Iris Setosa','Iris Versicolour','Iris Virginica','Location','NorthWest');
 
hold off;

Категория: Непроверенные учебные задания