帶邏輯鏈接功能的SVM迴歸？

Question

我使用的彈性網絡分類器需要概率輸出介於0和1之間的模型，我發現SVM比glmnet模型有更好的分類準確性（並不令人驚訝）。我知道e1071支持SV迴歸，有沒有辦法指定鏈接函數並獲得支持向量邏輯迴歸？謝謝。使用R 3.01，順便說一句。

Answer 1

而不是使用logit鏈接功能（我不知道你會怎麼做），我認爲你所需要做的就是在你的模型參數中設置probability=True。

試試這個例子：

library(e1071) 

data(iris) 

#convert this into a two class problem with 0's and 1's in our response 
subs_iris = subset(iris, Species != 'virginica') 
subs_iris$Species = ifelse(subs_iris$Species == 'setosa',0,1) 
attach(subs_iris) 

x <- subset(subs_iris, select = -Species) 
y <- Species 
model <- svm(x, as.factor(y), probability=T) 
(pred <- predict(model, x, probability=T)) 


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 

22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 

43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 
0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 

64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 

85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 

attr(,"probabilities") 
       0   1 
1 0.987105973 0.01289403 
2 0.981947817 0.01805218 
3 0.985437173 0.01456283 
4 0.982502592 0.01749741 
5 0.986736251 0.01326375 
6 0.985797388 0.01420261 
7 0.982548458 0.01745154 
8 0.986618794 0.01338121 
9 0.978261097 0.02173890 
10 0.983057145 0.01694286 
11 0.985273718 0.01472628 
12 0.985044518 0.01495548 
13 0.982882845 0.01711715 
14 0.977350638 0.02264936 
15 0.977353717 0.02264628 
16 0.977337824 0.02266218 
17 0.986633849 0.01336615 
18 0.987559915 0.01244008 
19 0.979015418 0.02098458 
20 0.986641340 0.01335866 
21 0.979223478 0.02077652 
22 0.987128146 0.01287185 
23 0.977319879 0.02268012 
24 0.977318096 0.02268190 
25 0.982043578 0.01795642 
26 0.977302751 0.02269725 
27 0.985124984 0.01487502 
28 0.986267482 0.01373252 
29 0.985675766 0.01432423 
30 0.983434135 0.01656587 
31 0.982646443 0.01735356 
32 0.979790608 0.02020939 
33 0.979143578 0.02085642 
34 0.981984290 0.01801571 
35 0.983332565 0.01666743 
36 0.985862990 0.01413701 
37 0.980783392 0.01921661 
38 0.983998736 0.01600126 
39 0.979876054 0.02
40 0.986253654 0.01374635 
41 0.987705405 0.01229459 
42 0.946960057 0.05303994 
43 0.978967777 0.02103222 
44 0.980641733 0.01935827 
45 0.984315643 0.01568436 
46 0.981331513 0.01866849 
47 0.985423052 0.01457695 
48 0.983645068 0.01635493 
49 0.986290826 0.01370917 
50 0.986350301 0.01364970 
51 0.029433804 0.97056620 
52 0.016430412 0.98356959 
53 0.024776288 0.97522371 
54 0.009426853 0.99057315 
55 0.016452416 0.98354758 
56 0.012704127 0.98729587 
57 0.019445907 0.98055409 
58 0.030085332 0.96991467 
59 0.018511622 0.98148838 
60 0.014208017 0.98579198 
61 0.029383457 0.97061654 
62 0.014964845 0.98503516 
63 0.026215448 0.97378455 
64 0.013069314 0.98693069 
65 0.022539574 0.97746043 
66 0.019169611 0.98083039 
67 0.018805048 0.98119495 
68 0.019024824 0.98097518 
69 0.029443333 0.97055667 
70 0.011480062 0.98851994 
71 0.026777215 0.97322278 
72 0.014014692 0.98598531 
73 0.021568195 0.97843180 
74 0.015284531 0.98471547 
75 0.015807559 0.98419244 
76 0.016950698 0.98304930 
77 0.025482614 0.97451739 
78 0.023218517 0.97678148 
79 0.012583780 0.98741622 
80 0.022016637 0.97798336 
81 0.010875445 0.98912455 
82 0.014092065 0.98590794 
83 0.013192719 0.98680728 
84 0.017883555 0.98211644 
85 0.024212461 0.97578754 
86 0.029433804 0.97056620 
87 0.018419284 0.98158072 
88 0.025893165 0.97410684 
89 0.022054824 0.97794518 
90 0.008828270 0.99117173 
91 0.010761677 0.98923832 
92 0.013793894 0.98620611 
93 0.011470407 0.98852959 
94 0.023992095 0.97600791 
95 0.010806582 0.98919342 
96 0.022895843 0.97710416 
97 0.015274431 0.98472557 
98 0.014182305 0.98581770 
99 0.029373597 0.97062640 
100 0.012725531 0.98727447 
Levels: 0 1 

說實話，我遇到了一些困難，在這裏訪問$probabilities屬性，但理論上你應該罰款只是看着每個記錄是在課堂上的概率「 1" 。