I have created a text classifier with RTextTools, but it does not perform that good. That is why I decided to tune the parameters gamma and c to improve the models' performance with the following code:
tc <- tune.control(cross=10)
modeltune <- tune.svm(x=tucontainer@training_matrix, y=tucontainer@training_codes, gamma= 10^(-6:1), cost = 10^(-1:1), tunecontrol=tc)
modeltune$performances
modeltune$bestmodel
However, I am getting the same result for every parameter combination and do not really have any idea, why.
gamma cost error dispersion
1 1e-06 0.1 0.4307769 0.01347589
2 1e-05 0.1 0.4307769 0.01347589
3 1e-04 0.1 0.4307769 0.01347589
4 1e-03 0.1 0.4307769 0.01347589
5 1e-02 0.1 0.4307769 0.01347589
6 1e-01 0.1 0.4307769 0.01347589
7 1e-06 1.0 0.4307769 0.01347589
8 1e-05 1.0 0.4307769 0.01347589
9 1e-04 1.0 0.4307769 0.01347589
10 1e-03 1.0 0.4307769 0.01347589
11 1e-02 1.0 0.4307769 0.01347589
12 1e-01 1.0 0.4307769 0.01347589
13 1e-06 10.0 0.4307769 0.01347589
14 1e-05 10.0 0.4307769 0.01347589
15 1e-04 10.0 0.4307769 0.01347589
16 1e-03 10.0 0.4307769 0.01347589
17 1e-02 10.0 0.4307769 0.01347589
18 1e-01 10.0 0.4307769 0.01347589
Does anyone know where the error is? Thanks in advance