1
2 import datafunc
3 import svm
4 import multi
5 import assess
6 import myio
7 import ker
8 import classifiers
9 import composite
10 import featsel
11 import modelSelection
12
13
14 d = datafunc.SparseCDataSet ('heartSparse.data')
15
16
17 print d
18
19
20 s = svm.SVM()
21
22
23 s.train(d)
24
25
26 r=s.cv(d)
27
28
29 print r
30
31
32 r.plotROC()
33
34
35 r.save('test.pyd')
36 r=myio.load('test.pyd')
37
38
39 d.attachKernel('polynomial')
40 rp=s.cv(d)
41
42
43 d.attachKernel('linear')
44 s = svm.SVM(ker.Polynomial())
45 r=s.cv(d)
46
47 d.attachKernel('linear')
48 knn = classifiers.KNNC()
49 r=knn.cv(d)
50
51 d = datafunc.DataSet ('heartSparse.data')
52 ridge = classifiers.RidgeRegression()
53 r=ridge.cv(d)
54
55
56 d = datafunc.SparseCDataSet('iris.data', labelsColumn = -1)
57 mc = multi.OneAgainstOne (svm.SVM())
58 r=mc.cv(d)
59
60 mc = multi.OneAgainstRest (svm.SVM())
61 r=mc.cv(d)
62
63 s = svm.SVM()
64 d = datafunc.SparseDataSet ('yeast.data', labelsColumn = 0)
65 d = datafunc.oneAgainstRest(d, '2')
66 r=s.cv(d)
67
68 m = composite.FeatureSelect (s, featsel.RFE())
69 r=m.cv(d, 3)
70
71 fs = featsel.FeatureScore ('golub')
72 f = featsel.Filter (fs, sigma = 2)
73 m = composite.FeatureSelect (s, f)
74 r=m.cv(d,3)
75
76 d = datafunc.SparseDataSet ('heart.data')
77 p = modelSelection.Param(svm.SVM(), 'C', [0.1, 1, 10, 100, 1000])
78 m = modelSelection.ModelSelector(p)
79 m.train(d)
80
81
82 d = datafunc.SparseDataSet ('heartSparse.data')
83 p = modelSelection.Param(classifiers.KNN(), 'k', [1,2,3,5,10,15])
84 m = modelSelection.ModelSelector(p)
85 m.train(d)
86
87
88
89 r = p.cv(d, numFolds = 10)
90 results = [r for r in p.cv(d, numFolds = 10)]
91 results = [r.successRate for r in p.cv(d, numFolds = 10)]
92
93 d = datafunc.SparseDataSet ('yeast.data', labelsColumn = 0)
94
95 d = datafunc.SparseDataSet ('yeast2.data', labelsColumn = 1)
96
97
98
99 from PyML import *
100
101 d = datafunc.VectorDataSet ('yeast3.data', labelsColumn = 1)
102
103 knn = classifiers.KNN()
104 results = knn.stratifiedCV(d)
105
106 p = modelSelection.Param(classifiers.KNN(), 'k', [1,2,3,5,10,15])
107
108 results = [r.successRate for r in p.stratifiedCV(d)]
109
110 m = modelSelection.ModelSelector(p)
111