For example the file heart_scale provided with the libsvm package. Do I need to make such files for training and testing? If yes how and on what basis?
Hi
first you take a decision that how many of all data is test and how many of them is test! of course you can use -v in libsvm and work with n-fold cross validation!
The format of training and testing data file is:
----------------------------------
: : ...
.
.
.
---------------------------------
Hi
first you take a decision that how many of all data is test and how many of them is test! of course you can use -v in libsvm and work with n-fold cross validation!
The format of training and testing data file is:
----------------------------------
: : ...
.
.
.
---------------------------------
Assume you have a feature set for a single object like:
010
011
001
You want to convert it to a training data in libsvm format. First of all, you should convert it to linear format like this:
010011001
Then you should add the class label to it. Assume the class for this data is +1:
+1 010011001
Now data should be represented in : format:
+1 0:0 1:1 2:0 3:0 4:1 5:1 6:0 7:0 8:1
You can omit 0 values:
+1 1:1 4:1 5:1 8:1
This is just for one single object. Finally you should create a file with each object's feature in separated lines.
Some sample train/test data:
http://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/binary.html
Thank you Mr.Mohsen Biglari. Each file should conatin exactly one object's features individually or the training/testing file would contain all the objects features but separated via lines?
Try reading:
https://sites.google.com/site/kittipat/libsvm_matlab
You need to divide your data into training subset and testing subset which mean you must to use two files; one for the training subset and the other for the testing subset. The most common way to divide your data into training subset and testing subset is by using k-fold cross validation with usually k = 5.
Good luck :)
@Hiba - How do we go about to split the sparse data into training and testing splits? Could you please display the code for doing so?
- Badges
- Science topic
- Similar topics
- Engineering
- Electrical Engineering
- Signal Processing
- Image Processing