IMUDataProcess/DatasetTestSVM.m at master · caunion/IMUDataProcess · GitHub

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
function do()
    svmpath = 'lib\SVM\libsvm-master\libsvm-master\windows';
    datapath= '..\dataset\';
    sbdtpath= '..\dataset\subset\';
    path(svmpath,path);
    path(datapath,path);
    path(sbdtpath,path);

    %startt = 1;
    %endt = 8;
    %[train,trainy,test,testy]=ReadDataSet(startt,endt);
    %[train,test] = loadFeature(startt,endt,trainy,testy);
    load('trainfeatures.mat');
    load('trainy.mat');
    train = trainfeatures;
    [data,label] = sample(train,trainy,2000);

%     test = testfeatures;
%     [data,label] = mergeData(train,trainy,test,testy);
%     [mdata] = LDAFeatureComp(data,label);
%     data = mdata;
    [train,trainy,test,testy] = DivideData(data,label,0.2);

    fprintf('data loaded!\n');
    [train_scale,test_scale] = scale(train,test);
    train = train_scale;
    test = test_scale;
    save('lda1t20_4000pertype.mat','train','trainy','test','testy');

    LinearKernelSVM(train_scale,trainy,test_scale,testy);
    GaussianKernelSVM(train_scale,trainy,test_scale,testy);
end
function [train_cale,test_scale] = scale(train,test)
    minimums = min([train;test], [], 1);
    ranges = max([train;test], [], 1) - minimums;
    train_cale = (train - repmat(minimums, size(train, 1), 1)) ./ repmat(ranges, size(train, 1), 1);
    test_scale = (test - repmat(minimums, size(test, 1), 1)) ./ repmat(ranges, size(test, 1), 1);
end

function [data,label] = sample(olddata,oldlabel,num_per_type)
    dims = size(olddata,2);
    labels = union(oldlabel,oldlabel);
    num = num_per_type * length(labels);
    data = zeros(num,dims);
    label = zeros(num,1);
    idx = 0;
    for i = 1 : length(labels)
        type = labels(i);
        I = find(oldlabel == type);
        sI = randsample(I,num_per_type);
        data(idx+1:idx+num_per_type,:) = olddata(sI,:);
        label(idx+1:idx+num_per_type,:) = oldlabel(sI,:);
        idx = idx + num_per_type;
    end
end
function [data,label] = mergeData(train,trainy,test,testy)
    if size(train,2) ~= size(test,2)
        return;
    end
    dim = size(train,2);

    labels = union(trainy,testy);
    data  = zeros(size(train,1) + size(test,1), dim);
    label = zeros(size(train,1) + size(test,1), 1);
    idx = 0;
    for i = 1 : length(labels)
        type = labels(i);
        traI = find(trainy == type);
        tesI = find(testy == type);
        num = length(traI) + length(tesI);
        data(idx + 1 : idx + num,:) = [train(traI,:);test(tesI,:)];
        label(idx + 1 : idx + num, :) = [trainy(traI,:);testy(tesI,:)];
        idx = idx + num;
    end
end
function [train,test] = loadFeature(startt,endt,trainy,testy)
    load('trainfeatures.mat');
    load('testfeatures.mat');
    startp =0;
    for i = 1: length(trainy)
        if trainy(i) ==startt
            startp = i;
            break;
        end
    end

    for endp = startp : length(trainy)
        if trainy(endp) == endt + 1
            endp = endp - 1;
            break;
        end
    end
    train = trainfeatures(startp:endp,:);

    startp =0;
    for i = 1: length(testy)
        if testy(i) ==startt
            startp = i;
            break;
        end
    end

    for endp = startp : length(testy)
        if testy(endp) == endt + 1
            endp = endp - 1;
            break;
        end
    end
    test  = testfeatures(startp:endp,:);
end

function LinearKernelSVM(train,trainy,test,testy)
    %linear kernel
    fprintf('Linear train begin\n');
    model_linear = svmtrain(trainy,train,'-t 0');
    fprintf('Linear train finished\n');
    save('model_linear.mat','model_linear');
    fprintf('Linear train saved, begin to test\n');
    [predicted_label_L, accuracy_L, prob_estimates_L] = svmpredict(testy, test, model_linear);
    disp(accuracy_L);
    save('accuracy_L.mat','accuracy_L');
    save('predicted_label_L.mat','predicted_label_L');
    fprintf('Linear test finished!\n\n');
end

function GaussianKernelSVM(train,trainy,test,testy)
    %Gaussian/RBF kernel
    fprintf('Gaussian/RBF kernel train begin\n');
    model_gaussian = svmtrain(trainy, train,'-t 2');
    fprintf('Gaussian/RBF train finished\n');
    [predicted_label_G, accuracy_G, prob_estimates_G] = svmpredict(testy, test, model_gaussian);
    disp(accuracy_G);
    save('accuracy_G.mat','accuracy_G');
    save('predicted_label_G.mat','predicted_label_G');
    save('model_gaussian.mat','model_gaussian');
    fprintf('Gaussian/RBF test finished!\n\n');
end