-
Notifications
You must be signed in to change notification settings - Fork 0
/
sequencialfsSVM_yu.m
281 lines (224 loc) · 7.84 KB
/
sequencialfsSVM_yu.m
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
function [fs,history]=sequencialfsSVM_yu(featureSet, modelo)
path(path,'svm')
s = char(featureSet);
fileNamed = ['results\',s,'_FeatureSelection.tex'];
arq = fopen(fileNamed,'w');
feat_folder = ['features\' featureSet '\'];
DS1 = {'101';'106';'108';'109';'112';'114';'115';'116';'118';'119';'122';'124';'201';'203';'205';'207';'208';'209';'215';'220';'223';'230'};
DS2 = {'100';'103';'105';'111';'113';'117';'121';'123';'200';'202';'210';'212';'213';'214';'219';'221';'222';'228';'231';'232';'233';'234'};
for j=1:size(DS1,1)
train_ds1(j) = str2double(DS1(j));
end
for j=1:size(DS1,1)
test_ds2(j) = str2double(DS2(j));
end
registers = {'101';'106';'108';'109';'112';'114';'115';'116';'118';'119';'122';'124';'201';'203';'205';'207';'208';'209';'215';'220';'223';'230'};
%registers = {'108';'109';'112';'114';};
% imprime a tabela de registros em arquvo .tex
%printRegisterTable(['features\' featureSet '\']);
fprintf('\n Gerando os modelos... \n')
tic
mdl=[];
scale_fac = [];
for k=1:size(registers,1) % numero de registros
fprintf('\n Gera modelo %d ... \n',k)
test_ds = [];
train_ds = [];
test_ds = str2double(registers(k));
count=1;
for j=1:size(registers,1)
if j ~= k
train_ds(count) = str2double(registers(j));
count = count +1;
end
end
%carrega os dados
%p1d = instancias
%p1t = label do treino
[p1d p1t p2d p2t] = loadDataAAMI(0,feat_folder,train_ds,test_ds);
%[p1d p1t] = undersampling_all_class(10, p1d,p1t);
[p1d, scale_factor] = mapminmax(p1d');
p2d = mapminmax('apply',p2d',scale_factor);
p1d = p1d';
p2d = p2d';
scale_fac{k}=scale_factor;
for t=2:size(p1t,2)
p1t(find(p1t(:,t)==1),t)=t;
end
p1t = sum(p1t');
p1t = p1t';
if nargin < 2 % apenas gera modelos se nenhum foi passado por parâmetro
cd svm
% cria os modelos (1 pra cada paciente)
% parametros para SVM
param_c=0.05;
param_g=1/(8 * size(p2t,2));
cmd = ['-w1 1 -w2 50 -w3 10 -c ',num2str(param_c),' -g ',num2str(param_g)];
mdl{k} = svmtrain(p1t,p1d,cmd);
cd ..
end
end
%{
% modelo de DS1 para ser aplicado a DS2
[p1d p1t p2d p2t] = loadDataAAMI(0,feat_folder,train_ds,test_ds);
%[p1d p1t] = undersampling_all_class(10, p1d,p1t);
[p1d, scale_factor] = mapminmax(p1d');
p2d = mapminmax('apply',p2d',scale_factor);
p1d = p1d';
p2d = p2d';
for t=2:size(p1t,2)
p1t(find(p1t(:,t)==1),t)=t;
end
p1t = sum(p1t');
p1t = p1t';
cd svm
fprintf('\n Gera modelo DS1 ... \n')
mdl{k+1}=svmtrain(p1t,p1d,cmd);
cd ..
%}
if nargin < 2 % apenas gera modelos se nenhum foi passado por parâmetro
save mdl
else
mdl = modelo;
end
toc
fprintf('\n passo 1. \n')
%vetor para indices das caracteristicas que vao ficar
keep = zeros(1,size(p2d,2));
keep(1,1:9)=1; % inicializa com as nove features de yu
sweep = zeros(1,size(p2d,2));
best_in = [];
best_acc = [];
best_acc_mean = 0;
%começo com 1 caraceristica
fprintf(arq,'------------------------------------------------\n');
fprintf(arq,'\nBusca para frente: \n');
fprintf('------------------------------------------------\n');
fprintf('\nBusca para frente: \n');
acc = zeros(size(registers,1),size(p2d,2));
%executa o processo de acordo com o número de caracteristicas
for i=1:size(p2d,2)
fprintf(arq,'\n Passo %d ... \n',i)
fprintf('\n Passo %d ... \n',i)
tic
%varre para todas as caracteristicas
for j=1:size(p1d,2)
if keep(1,j) ~= 1 || best_acc_mean == 0
sweep = keep;
sweep(1,j) = 1;
% end
%testa para todos os pacientes
for k=1:size(registers,1)
train_ds = [];
count=1;
for ll=1:size(registers,1)
if ll ~= k
train_ds(count) = str2double(registers(ll));
count = count +1;
end
end
test_ds = str2double(registers(k));
[p1d p1t p2d p2t] = loadDataAAMI(0,feat_folder,train_ds,test_ds);
%[p1d p1t p2d p2t] = loadDataAAMI(0,feat_folder,[],test_ds);
%coloca valor médio nas colunas em que sweep estava zero
p2d(:,find(sweep==0)) = repmat(mean(p1d(:,find(sweep==0))),size(p2d,1),1 );
%[p1d, scale_factor] = mapminmax(p1d');
p2d = mapminmax('apply',p2d',scale_fac{k});
%p1d = p1d';
p2d = p2d';
%{
for t=2:size(p1t,2)
p1t(find(p1t(:,t)==1),t)=t;
end
p1t = sum(p1t');
p1t = p1t';
%}
for t=2:size(p2t,2)
p2t(find(p2t(:,t)==1),t)=t;
end
p2t = sum(p2t');
p2t = p2t';
acc(k,j) = evaluation(p2d,p2t,mdl{k});
%fprintf('Acuracia para registro %d %x')
end
else
acc(k,j)=best_acc_mean;
end
end
% acha a maior acurácia média para todos os pacientes
acc_mean = mean(acc,1);
index = find(acc_mean == max(acc_mean));
if(acc_mean(index(1)) > best_acc_mean)
best_acc_mean = acc_mean(index(1));
% atualiza keep
fprintf(arq,'\n **atualiza fs - Melhor configuração da rodada (acc medio=%f)\n fs =',acc_mean(index(1)));
fprintf('\n **atualiza fs - Melhor configuração da rodada (acc medio=%f)\n fs =',acc_mean(index(1)));
keep(1,index(1))=1; % mantem a melhor configuração
for uu=1:size(keep,2)
fprintf(arq,'%1d ',keep(1,uu));
end
for uu=1:size(keep,2)
fprintf('%1d ',keep(1,uu));
end
else
fprintf(arq,'\n Inclusão de novas caracteristicas não melhorou o resultado!\n Processo terminado. (acc medio=%f)\n fs =',acc_mean(index(1)));
fprintf('\n Inclusão de novas caracteristicas não melhorou o resultado!\n Processo terminado. (acc medio=%f)\n fs =',acc_mean(index(1)));
for uu=1:size(keep,2)
fprintf(arq,'%1d ',keep(1,uu));
end
for uu=1:size(keep,2)
fprintf('%1d ',keep(1,uu));
end
i=size(p2d,2); % força o término
%fclose(arq);
break;
break;
end
toc
end
fprintf(arq,'\n\n Melhor configuração final \n');
fprintf(arq,'fs = ');
for uu=1:size(keep,2)
fprintf(arq,'%1d ',keep(1,uu));
end
fprintf('\n\n Melhor configuração final \n');
fprintf('fs = ');
for uu=1:size(keep,2)
fprintf('%1d ',keep(1,uu));
end
fprintf('\n');
fclose(arq);
end
function cv =evaluation(inst, label, mdl)
wa = @(e1,e2,e3,s1,s2,s3)(1-(e1*1/s1+e2*1/s2+e3*1/s3)/3)*100; %weighted accuracy
if max(unique(label)) <=1
label(1)=2;
end
cp = classperf(label);
[class] = svmpredict(label,inst,mdl);
classperf(cp,class);
if(size(cp.ClassLabels,1) >= 3)
cv = wa(cp.errorDistributionByClass(1), ...
cp.errorDistributionByClass(2), ...
cp.errorDistributionByClass(3), ...
cp.SampleDistributionByClass(1),...
cp.SampleDistributionByClass(2),...
cp.SampleDistributionByClass(3));
else
if(size(cp.ClassLabels,1) >= 2)
cv = wa(cp.errorDistributionByClass(1), ...
cp.errorDistributionByClass(2), ...
0, ...
cp.SampleDistributionByClass(1),...
cp.SampleDistributionByClass(2),...
1);
else
cv = wa(cp.errorDistributionByClass(1), ...
0, ...
0, ...
cp.SampleDistributionByClass(1),...
1,...
1);
end
end
end